Last month I noted the extremely long time it
takes Linux to delete large files. It took nearly 30 seconds elapsed time to delete a 190
GB file, while FreeBSD does it in a few
milliseconds. One hypothesis was that FreeBSD's speed was only due to soft updates, and
there's some reason to believe that. Today was the first of the month and thus the time for
my monthly level 0 dump of my file systems. I back up to disk, so I first needed to make
space by deleting a number of large files, 36 GB in total. They're on a separate file
system, so I was able to umount it, reset soft updates, mount it again and delete the
files. I then set soft updates again, copied a similar backup and deleted it. Here the
results:
No soft updates:
real 0m2.186s
user 0m0.006s
sys 0m0.800s
Soft updates:
real 0m0.127s
user 0m0.000s
sys 0m0.002s
So yes, it's much faster with soft updates than without. Interestingly, the processor time
also drops to almost 0; I suspect most of the time is in non-process context,
which time(1) can't measure. But the time without soft updates is about 0.062s per
GB. For Linux (XFS) it's 0.63s per GB, 10
times as much. My (possibly inaccurate) recollection is that it was even slower
with ext3. There's still something puzzling
about the figures.
There are differences in the scenarios: I think my FreeBSD box is faster than the Linux box,
though not much, and I was deleting a number of smaller files. But I don't think that that
is particularly relevant. It's clear that the CPU time is not an issue in the equation
(indeed, the slower Linux box used only 1.112s CPU time to delete 190 GB, while my faster
FreeBSD box needed 0.800s to delete the 36 GB). The disks are effectively the same speed.
I can't see any reason why it should be faster to delete a number of files than a single
file of the same size; I'd expect it to be slower due to other overhead.
While walking around the garden, discovered that we had forgotten a tree to chop down
yesterday, presumably a River red
gum. Grabbed a handsaw and chopped it down:
Also spent some time removing the inordinate amount of grass that had popped up there. We
should try to stay more on top of the weeding.
Hugin has a number of poorly
documented features, one of which is the plethora of choices on the Stitcher page:
It occurred to me that it would be nice to merge the “before” and
“after” photos of the gum tree removal to get an exact view of the change. That
should be possible with programs like align_image_stack, but my previous attempts with it proved to be less than successful. Some of
the choices on the Stitcher page looked like they would do what I want.
If they do, I couldn't find it. But some of the other choices looked interesting,
particularly “Merged and blended panorama”. Currently I do this in two steps:
first I create HDR images of each view, and then I stitch the HDR images. It would be nice
to do it automatically. Tried that with one of last
Saturday's panoramas, without success:
enblend: an exception occured
Mask is entirely black, but white image was not identified as redundant.
gmake: *** [p.tif] Error 1
Maybe I can drag out some more information about the topic.
Out walking with Nemo today. You can tell that
elections are coming; people are investing in infrastructure:
It's not clear what that sign is supposed to mean, but I think the money would have been
better spent removing flammable debris from the roadside. Of course, silly signs are
popular round here; only 150 m or so further down the road there's another one that has been
there for a while:
That's on one of the many entrances to
the Dereel lagoon. There's no other sign
there, and you can go off in any direction from there. There's no other similar sign
elsewhere. So where exactly are dogs and horses prohibited?
Our Acer 530 data
projector has just passed the 3000 hour lamp life limit, and almost immediately the light
output has dropped dramatically—we could barely see anything any more. I bought it
at a time when price was the primary objective, and it's
not much of a projector—even Acer has
completely forgotten about it—and I had never intended to replace the lamp. But what
do we do now? In the short term it's simple: use the old Panasonic instead. The colours
are terrible, but the image is bright enough. And then buy the 1080p projector I've been
thinking about for years? I think I've changed my mind: nearly all the video we look at is
720p or less, and it's to be expected that a 720p projector will render 720p content better
than a 1080p one. They're also cheaper and tend to be brighter. So tomorrow it'll be time
to start another search.
The weather is dryer now, but not dry enough for Ron of Enfield Mini-Diggers, so he put
things off—again—until Monday. CJ showed up, however, and spent the whole day
working on the new dog run for Nemo. It proved to
take longer than expected, and by evening we still weren't done. Hopefully Tuesday now.
Over the last few days I have gradually come to the conclusion: I need a 720p projector
(1280x720), not a 1080p (1920x1080), since that's what the TV stations broadcast. Today
spent nearly the whole day looking for a suitable projector. What I found was:
-
As I had expected when I bought the last projector,
the standard projector now is 1080p.
-
Those 720p projectors still on the market are either old (such as the Panasonic
PT-AX200E) or not really intended for home theatre use (such as the Epson TW450, which has a real
resolution of 1280x800 and significantly higher noise levels).
-
A limiting factor for my particular application is the size of the room: it's not very
wide, and there would only be about 3 m between the projector and the screen. Only some
projectors can project an adequately sized image at that distance.
But which ones? I know my old Panasonic PT-AE700E can, and I downloaded the manual for
the PT-AX200E, which confirmed that it had exactly the same parameters (probably the
same lens as my 5 year old PT-AE700E). But what of the others? The documentation on
the web sites has hit a new low. For example, the “spec
sheet” for the Epson TW450 states:
| SCREEN SIZE (PROJECTED DISTANCE) |
33 to 318 inch [0.9 to 9.0m] (Zoom: Tele), |
|
|
33 to 317 inch [1.08 to 10.5m] (Zoom: Wide)
|
What on earth does that mean? Our best guess on IRC is that the inches measurements are
the screen diagonal and the metric units are the distance, which would mean that you can
get a 33" display at distances between 90 cm and 108 cm (convenient metric values
that presumably just happen to correspond to 3' and 3' 6"). What kind of
mind can create that? A mix of units, inadequate description of the meaning, and a
representation that requires significant calculations to work out what the size would be
at a given distance. If I take that trouble, I discover a maximum diagonal of 2.8 m at
3 m distance. The Panasonics have 2.5 m, so that would seem OK. But is my guess
correct?
This would be terrible enough if only Epson did it, but I've seen other manufacturers do
it as well. Sure, you don't need a university degree to become a web master (I suspect
it might even be a disadvantage), but this complete mess makes me wonder if
mind-altering substances are a requirement. Wouldn't it be easier to say that at
distance n the diagonal can be between 0.79 n and 0.92 n?
-
Projector bulb life is very variable. The Epsons offer up to 4000 hours, while the
Panasonic PT-AX200E has a hard limit of 2000 hours, after which it refuses to work (OK,
you can reset the life timer, but they warn about exploding bulbs, just as they did for
my old projector).
-
Warranties are limited. Panasonic only has 1 year (particularly significant in view of
the damage that happened to my old projector after about 18 months), and there are some
interesting
clauses: not more than 4 hours continuous use. This relates to the optical block,
which is where my projector was damaged, so I assume the Panasonics have a weakness in
this area. Epson has 2 years' warranty, and Sanyo has 3 (which according to Ausmedia also includes the globe), though their web
site doesn't mention anything about the warranty terms.
By evening I still hadn't got much further. Did some experimentation and decided that 1080p
projectors don't seem to have difficulties displaying 720p, so it might make more sense to
bite the bullet and get an up-to-date projector than buy something that is no longer even
state of the art.
Yet another issue with Panasonic became apparent in the evening. The X display didn't work;
instead of 1280x720, I only got 640x480. Further investigation showed that the projector
doesn't supply EDID information, and without mode lines X assumes that it can't use higher
sync frequencies than VGA. This problem didn't hit us yesterday because the Acer projector
was connected when I started X. Worked around that by reconnecting the Acer projector,
starting X and then swapping projectors. Here the relevant diffs
from /var/log/Xorg.0.log:
--- Xorg.0.log.lowres 2010-07-02 21:22:39.000000000 +1000
+++ Xorg.0.log.hires 2010-07-02 21:25:19.000000000 +1000
@@ -413,7 +413,6 @@
(**) NVIDIA(0): Enabling RENDER acceleration
(II) NVIDIA(0): Support for GLX with the Damage and Composite X extensions is
(II) NVIDIA(0): enabled.
-(WW) NVIDIA(GPU-0): Unable to read EDID for display device CRT-0
(II) NVIDIA(0): NVIDIA GPU Quadro NVS 210S / NVIDIA GeForce 6150LE (C51) at
(II) NVIDIA(0): PCI:0:5:0 (GPU-0)
(--) NVIDIA(0): Memory: 131072 kBytes
@@ -421,17 +420,17 @@
(--) NVIDIA(0): Interlaced video modes are supported on this GPU
(--) NVIDIA(0): Connected display device(s) on Quadro NVS 210S / NVIDIA
(--) NVIDIA(0): GeForce 6150LE at PCI:0:5:0:
-(--) NVIDIA(0): CRT-0
-(--) NVIDIA(0): CRT-0: 350.0 MHz maximum pixel clock
+(--) NVIDIA(0): Acer PH530 (CRT-0)
+(--) NVIDIA(0): Acer PH530 (CRT-0): 350.0 MHz maximum pixel clock
(II) NVIDIA(0): Assigned Display Device: CRT-0
-(WW) NVIDIA(0): No valid modes for "1280x720"; removing.
-(WW) NVIDIA(0): No valid modes for "800x600"; removing.
(II) NVIDIA(0): Validated modes:
+(II) NVIDIA(0): "1280x720"
+(II) NVIDIA(0): "800x600"
(II) NVIDIA(0): "640x480"
-(II) NVIDIA(0): Virtual screen size determined to be 640 x 480
-(WW) NVIDIA(0): Unable to get display device CRT-0's EDID; cannot compute DPI
-(WW) NVIDIA(0): from CRT-0's EDID.
-(**) NVIDIA(0): DPI set to (40, 54); computed from "DisplaySize" Monitor
+(II) NVIDIA(0): Virtual screen size determined to be 1280 x 720
+(WW) NVIDIA(0): Acer PH530 (CRT-0)'s EDID does not contain a maximum image
+(WW) NVIDIA(0): size; cannot compute DPI from Acer PH530 (CRT-0)'s EDID.
+(**) NVIDIA(0): DPI set to (81, 81); computed from "DisplaySize" Monitor
(**) NVIDIA(0): section option
(==) NVIDIA(0): Enabling 32-bit ARGB GLX visuals.
(--) Depth 24 pixmap format is 32 bpp
In addition, I was reminded the hard way that the projector can't synchronize to the
frequencies that FreeBSD uses for the console,
which is slightly higher than normal VGA. I doubt that they have invested much in the newer
720p modes. There are clearly enough reasons not to buy a PT-AX200E: limited warranty, poor
compatibility, previous reliability issues, low lamp life and relatively high price. The
cheapest price I found for one is $1970; for $2170 I can buy a Sanyo PLV-Z700 1080p unit
with three years' warranty and a short-focus lens slightly better even than the Panasonic.
About the only down side (apart from the fact that I didn't really want a 1080p) is that the
lamp is relatively dim at 1200 lumens. But that's still brighter than the Panasonic I have
now (1000 lumens).
The Ballarat Linux Users Group is interested
in hosting linux.conf.au in 2012.
That's quite a challenge: so far it has only been hosted in big cities. The smallest in
Australia was Hobart, with a population of 220,000, and that proved a problem at the time.
Ballarat has less than 80,000 inhabitants. That reflects itself in the number of people
available to organize the conference, and also in the range of venues which we could use.
This evening Josh Stewart organized a first meeting at the Thai Sala Pavilion Restaurant in Sturt
Street. As recently at my weather station
presentation, the size of the town was evident: we had only 5 people, one of whom
still needs to develop the sense of responsibility needed for such an event:
Still, I think we should be able to manage it.
Into Ballarat again today for the second week of Nemo's training. Socialisation, reinforcement of what he learnt last week, lots of questions,
including control and pecking order, and relatively little new stuff. We're making good
progress, Ian says, but young puppies are still quite a handful.
Ian doesn't miss much. He saw the tiny scratch that Nemo got from the neighbour's dog and asked what happened. When we told him,
he said “what kind of dog?”. I told him it was a golden retriever, and he said
“He'll remember. One day he'll be big, and that dog had better watch out”. I
wonder what will happen there.
We're still playing around with the navigator, which from time to time asks us to go ways
that we wouldn't normally take. Today we tried some of them; it's not clear whether they're
better, but it's probably worth trying it out.
One of the problems is data accuracy. On the way into town, at one point we were told to
turn second left. Only one street. Drove past and were told “Recalculating”,
in other words, “you blew it”.
On the way home we stopped off
in Napoleons at the little roadside
plant sales place. Didn't buy anything, and since we were about 50 m from the main road, we
should have turned around and continued on that road. But the navigator wanted us to continue over a kilometre and take a right turn into this road:
It's not clear why: it's clearly longer than turning back. Possibly it had other tricks up
its sleeve. Clearly a navigator can only be as good as the map data, and this is deficient.
But it reminded me of a recent Microsoft advertisement in issue 10/2010 of c't, with a claim:
CODE BEWEGT DIE WELT
In English, that's “Code moves the world”, and at the time I marvelled at the
stupidity, though it's in line with my experience with Microsoft MUAs. Looking for it now, it's no longer there, not even
on the web site; Google draws a blank. So maybe they have realised how stupid the idea is
and removed it again. The only hit I got in English was more sensible:
... the idea is that content and not just code moves the world
Clearly this experience helps relativize that claim.
Across the road to Lee and Ray Nottle in the evening; they were celebrating a Moroccan
feast, though we were a few people short: the Riebelings from Bendigo didn't make it, so the only guests were Chris Yeardley and ourselves, faced
with an enormous spread of food:
It seems that Lee spent some time in Israel in her youth, and learnt a lot of Middle Eastern
and North African food. Ate far too much.
It's been a while since we've had significant problems with kangaroos, but they were back
today. Chased them off; hopefully they'll stay away.
The Bureau of Meteorology has been even less
accurate than usual with the weather forecast. They had forecast -1° for this morning, but
in fact the overnight low was 6.8°. It still wasn't exactly inviting, but Ron of Enfield
Mini-Diggers is coming tomorrow (maybe), so spent some time preparing for that, in the
process finding even more spring bulbs. I really need to keep better track of what I plant
and where. These ones went in around the bird bath, which is now getting saturated.
The kangaroos are really coming in in force again now. We had to chase them off twice
today. Hopefully they'll stay away; they can't be hungry right now.
While looking for something completely different today, came across RFC 681, dated 18 March 1975. It appears to
be the earliest use of UNIX (Fifth Edition) in an (ARPA) network environment. The most
interesting thing from my point of view is that they use the standard file system interface
to talk to the network. It's a pity it doesn't give more detail.
The other day I did some comparisons of the quality of a 720p broadcast stream displayed at
natural resolution and also scaled up to 1080p. We discussed it on IRC today, and I took
some screen shots. The result? What I sort of knew already: the quality of Australian HDTV
isn't really very good. We're using MPEG-2, which makes relatively inefficient use of
bandwidth. What's the bandwidth? According to the transcoders, it's 90 Mb/s, or 11.25
MB/s. But that would give a recording size of 40 GB per hour. In fact, a typical 35 minute
recording uses 2358107876 bytes, or 1.123 MB/s—a number suspiciously close to 10% of
the nominal rate. The results are clear: digitalization artefacts, particularly in scenes
with motion in them. The following two images are 720p and 1080p, and they really need to
be looked at full size:
There are many artefacts, but the ones around the scoreboard are particularly obvious.
Clearly any loss in the scaling pales by comparison. I hear from friends overseas using
MPEG-4 and H.264 that the results are much better.
On a different topic, updated my X configuration to work with the Panasonic projector. That
was easy: I had commented out the entry, so I just needed to change it. But the image
quality got much worse. Further examination showed that the mode line I had chosen
generated a different horizontal frequency. Here the “old” (Acer) and the
“new” (Panasonic):
ModeLine "1280x720" 99 1280 1324 1420 1656 720 725 728 772 +hsync +vsync
ModeLine "1280x720" 76 1280 1324 1420 1656 720 725 728 772 +hsync +vsync
The first generates a horizontal frequency of 59.8 kHz, and the latter one of 45.9 kHz.
Both are in spec, but it seems that the Panasonic doesn't handle them equally well. After
the change, the image quality was much improved.
Another thing I've changed, on the recording side, is to stop using the
broadcast EPG. The quality is much worse
than the Shepherd guide was, and since then
Shepherd has improved again: there's clearly a lot going on behind the scenes. The EPG
seems to be yet another unloved stepchild. Without sensible categories, it's too difficult
to find what you're looking for.
The experiment with salsa ranchera last month were rather disappointing, so
today I tried an experiment with typical Mexican ingredients: apart from tomatoes, onions
and garlic, also put in some Chile
ancho, cumminseed and oregano. The
result? My salsa falsa. Put some on
some hamburgers, where it didn't taste bad, but I don't think that's the best use for it.
Round here there's nothing unusual about getting an ECONNRESET error loading web
pages, and when I got one with Wikipedia today I just assumed that it was the normal flakiness of my satellite link. But this time
it stayed, and it proved to be a power failure in their primary data centre. The interesting thing is that the
details are all available online at URLs like http://ganglia.wikimedia.org/ and http://nagios.wikimedia.org/, though the latter seems a
little behind the times. Spent some time watching that until we realised that the system
was back up, but Nagios still claimed lots of down systems.
So: I've decided on a Sanyo PLV-Z700 projector, for the reasons I've described in some detail. Where to
buy it? According to StaticIce, it's
available at prices between $2255 and $3300. But I got a lot of information from Ausmedia, who don't figure on that list. They have
copious information and well-hidden prices, which suggest that they'll be expensive. They
also don't mention any pick-up facility, which suggests fun with the delivery company.
Checked the price anyway, and to my surprise it was only $2170. Even with delivery, it was
cheaper than the next cheapest. Sent off an “email” (why do webmasters call
HTML forms “email”?) and asked if I could pick it up, and where. Got an answer
in 12 minutes (at 8:34) telling me that pickup was fine, and that they're in Station St., Box Hill, an eastern suburb of Melbourne. All looks very positive;
transferred the money, and I'll be in town tomorrow to pick it up.
Call from Enfield Mini-diggers today telling me that the ground was still too wet, and that
they'd come tomorrow. I suspect that this is another way of saying they haven't finished
the last job yet. And tomorrow's out of the question, because I'll be in Melbourne. So the
current date is Wednesday.
Did another hour or so's work in the garden, cleaning out much of
the Carpobrotus (“Pig
face”, or as we prefer to call it, “pig's ear”) from the northern bed. If
I could only stick to this much every day, I think I'd be on top of it.
Off to Melbourne today with Chris. We had intended to buy a USB DVB-T tuner, which is a
compatibility minefield, so did some research: what MSY offers, what LinuxTV.org told me, and what MSY really had in stock, so started printing out some
information. To annoy me, the printer started jamming in the paper feed, and I ended up not
being able to print the LinuxTV page.
Off to Melbourne by the “short” route that the navigator had offered me, through
the Brisbane Ranges. It's close
to a route that I have already taken, but not quite the same. It took us
through Werribee, and got us to MSY in
Brooklyn in just under 90 minutes. Had to wait longer than usual; there were only two
people in front of us, but they both bumbled and mumbled without buying anything in the end.
Got the tuner and a new DVD burner—only $34 nowadays—and on to Reinhard Bahlo's
favourite German Butcher, Fleischer's (nomen est
omen). They have a good choice of just about everything except what I wanted, but some
of the other stuff looked good, so I bought a little more, and Chris stocked up on lots
of Aufschnitt (and no, you can't
translate the word, even if Wikipedia thinks you can).
Then on to Ausmedia to pick up the
projector. They checked, but the money hadn't arrived. Spent quite some time waiting, and
in the end they suggested that I log in to my web banking account and check if it had been
deducted from my account. It hadn't; one of these horrible things about ANZ web banking is that they present a full screen (in
this case, 1920x1200) page which doesn't even display all the information, and then hide a
little “Confirm” button outside. Not for the first time, I had forgotten.
So they suggested I transfer the money again, and they would give me the projector. Did
that and got the projector. That's really so different from other companies; they'd
be well within their rights to say “Sorry, mate, the money's not here. We'll contact
you when it is, and you can come and pick it up”. Or I could have paid by credit card
(1% more). I'm quite positively impressed.
Ausmedia is in Box Hill, and
it was (somewhat past) lunchtime, so to the station complex for lunch, which was plentiful
but not as good as I recall from last time.
On to IKEA with a list of about 8 things to
get; after a lot of frustrating searching, I only got two, and I'm not convinced that
Yvonne will like one of them. What a pain.
Finally off, at a snail's pace, to Trinity College for the LUV meeting. It
took 45 minutes for a journey of 6.7 km, but we still arrived 1¼ hours before the meeting started. Walked around the
university campus again, not for the first time, and tried to locate the kindergarten where
I started my formal education in March 1952. Found a similar looking building to the north
of the oval, now the home of the Melbourne University
Mountaineering Club, but it looks bigger than I recall; given how much I've grown
since then, it should look smaller.
To the meeting, which was supposed to be about the use of GIMP in processing digital photos, but it barely touched on
GIMP in the nearly 2 hours it went on. Quite an interactive discussion, with questions like
“can digital cameras take photos in this light?”, clearly a yes:
That was taken with my old Nikon L1 camera steadied on the seat, thus the unfortunate
angle. There were a lot of other discussions which seemed off-topic to me (and Chris),
including a long discussion of analogue photography. What I came away with is:
-
What I saw of GIMP in no way made it easier for me to make friends with it. Still far
too much mouse pushing.
-
The things that Stewart demonstrated ran much faster than on my machine? Why?
Mine isn't the fastest in the world, but this was a laptop. I can't imagine that it
would be more than double as fast as mine.
-
There's a program called f-spot, which
for some reason I keep wanting to call f-point, that organizes photos. Stewart
uses it a lot, but it seems to me to be one of these things that want to take over the
entire organization of the photo collection. At least the way Stewart had it, that
includes showing things backwards. On the
other hand, it doesn't offer you the opportunity to put a title to a photo (“who
needs that?”). I suppose I should take a look at it, but I don't see myself using
it.
-
Another program is darktable, apparently a word play on Light Table, a term I haven't heard before (though I have one). It has other
æsthetic issues, like reverse video, but it also seems to have quite a few interesting
functions. I should try using it sometime.
We didn't finish until 21:00, and decided not to go to dinner with the others; instead off
to Footscray to look for a Vietnamese
restaurant, not helped by the fact that we ended up on the inner lanes of Flemington Road,
where it's almost impossible to turn off. Found a restaurant, had some food (I had
a Phở that wasn't bad, but which I wouldn't
choose again if I returned), and then off home, getting home shortly before midnight.
Somehow the effort wasn't worth it.
The proof of the pudding is in the eating, and so it is
for Aufschnitt. Tried out the
coarse Leberwurst that I bought
yesterday. Unfortunately, it's very good. That means that we'll have to make the
journey there more often, and it's long.
I've been doing much more thinking about panorama hardware than I have written here, and
a few days ago I invested my first money in additional hardware: $0.72 for two 3/8"
screws to enable me to attach my panorama head to my otherwise useless ball head. They
needed some trimming to fit:
Normal tripod screws are ¼", so I had to drill out the hole in the tripod plate to
accommodate the larger screw:
Now I can mount the pan head on the ball head. Level the ball head and I have a horizontal
plane on which to pan:
I'll see how well that works on Saturday, but so far it looks good.
Spent some time setting up the new Sanyo PLV-Z700 projector
today. It's much bigger than the old ones. Here the Acer, the Panasonic and the
Sanyo from top to bottom. The backs of each projector line up.
Things started off well: plug in, turn on, and it displayed a perfect image—in
1280x720, because that's what my X configuration said. Getting it to work correctly in
1920x1080 was a very different matter. Decades ago I wrote the book when it comes to setting up X mode
lines, but in those days displays were invariably CRTs. Times have changed, and projectors
seem to have a lot of idiosyncrasies that even LCD panels don't have. Today X ignored all
my mode lines with the message:
(WW) NVIDIA(0): No valid modes for "1920x1080"; removing.
Why? I had a mode line:
ModeLine "1920x1080" 150 1920 2048 2248 2576 1080 1083 1088 1120 +hsync +vsync
Normally X tells you why the mode lines aren't valid (“Horizontal sync out of
range” or some such) but here there was no explanation. Checked all the parameters,
all the messages, and made no progress. Finally ran X -configure and took the
generated xorg.conf file. That worked, but the picture was positioned far to much to
the right. xvidtune didn't help much; I gradually moved the image to the left, but
it got dimmer in the process, something that looked worryingly like the way to kill a
monitor in the bad old days.
The handbook (conveniently not included in the 220 page book included with the projector;
that's a “Quick Reference Guide” and is 20 pages × 11 languages) tells me that
it will basically accept any signal up to 100 MHz dot clock, but it gives as the 1080p
frequencies 67.5 kHz/60 Hz (for USA and other compatible countries) and 56.25 kHz/50 Hz (for
the rest of the world). The absolute theoretical minimum dot clock for 1920 dots at 67.5
kHz is 130 MHz, so it's not clear what that statement can mean. Ignored the dot clock
limitation and tried to set the horizontal and vertical frequencies, not without some
surprises:
The information display on the left doesn't seem to even come close to the values
that xvidtune and I expected: this image was at 56.31 kHz/50.28 Hz, but the projector
claims it's 52.1 kHz/46.6 Hz. And to get that I have a pixel clock of 132 MHz, way outside
the claimed range. Played around for a while and got an acceptable image, but the image
quality looks like it could do with improvement.
Spent some time looking for alternative sources of mode lines. The first one started with a reference to this diary, as did a one more specific to the projector. Gave up and considered the Microsoft space.
How do you convert data in a Microsoft “driver” to a mode line? The simplest
way seems to be to read the data out of the GPU registers, if that's possible. Investigated
a little, and on a suggestion looked at the CD that came with the projector. It
contained only the manual—no software at all. That makes sense—I never
understood why Microsoft space devices always require a “driver”—but I
think that's the first time I've seen a display device that comes without one.
On recollection, it's not that surprising. Projectors and digital TV displays take
standardized signals, for example from DVD players or TV broadcast, so the timing must be
standardized. All I need to do is to find the standard. In the meantime, found a reference
to a 1080p mode line for a different display in the MythTV modeline database.
ModeLine "1920x1080" 132 1920 2048 2248 2344 1080 1083 1088 1120 +hsync +vsync
It looked wrong: the sync pulse was too close to the end of the display, which should have
pulled the display to the right, but when I tried it, it worked relatively well. So I now
have the thing running, and can spend more time tuning to perfection.
And what's it like? The lens doesn't seem as good as the Panasonic: there's some chromatic
aberration, which is visible in the photo above. But the colours are an order of magnitude
better than the Acer, and the brightness is good, far more than I expected for a projector
with only 1200 lumens. Some time I must measure the current brightness of all three
projectors.
While paying for the projector yesterday, I entered my web banking password on somebody
else's computer. There's no reason to believe that they have sniffed it, but there's also
no reason to take the chance, so for the first time in a long time I went to change it.
That wasn't easy. First Yvonne suggested
“Nemolino” (her diminutive of “Nemo”), but that was rejected because it contained no digits. So I
tried something else, which was rejected because it contained non-alphanumeric characters:
Your password should contain only letters and numbers. Your password must also contain at
least one number and one letter. Please re-enter your password to meet these security
requirements for your finance information.
This is presumably the kind of stupidity that results in passwords like Abc123.
Finally conformed with a password reflecting my opinion of the intelligence of the people
who these restrictions. When will they ever learn?
SkyMesh is taking forever to resolve my
satellite link problems. Each message to support takes two weeks to get a response to, and
I'm no longer prepared to accept the claims that the problem is a defect in the modem. They
pointed to the “power on” events as “evidence” for that, but they
haven't been able to prove it. Things are not getting any better, so sent off a message to
Paul Rees, the manager, who in the past has been helpful, and suggested that if he thinks
it's the problem, they should install a new modem and prove it. Got a reply back
(Microsoft-style, appending the message rather than answering it, and thus missing several
points) telling me that they didn't want to spend the money, and that I could get a modem
elsewhere if I want. He also appended the log of the most recent outages, only one of which
was a “power on” (and that was me power cycling it). It looks as if I'm going
to have to complain to the DBCDE or the
TIO.
Read an interesting statement on a mailing list today:
It seems that pre-meeting dinners tend to be at Asian restaurants (going by the UK
definition which includes India).
The UK definition includes India in Asia? I'd very much hope so. But I discussed with a
number of people, and here in Australia it seems that for them “Asian” refers
particularly to East Asians
(China,
Korea,
Japan, maybe what used to be
called Indochina). In the UK, as the
writer observed, they include India. What
about Iran? It seems that nowadays people
tend to use the term “Asian” to mean “coming from one of those Asian
countries most represented here”. If that's the case, I
suppose Việt Nam is also Asia. But what
about Laos?
Finally the weather was right, and Ron of Enfield Mini-Diggers showed up to dig the pond.
That was only a small part of the work, as it turned out: the soil there is so soft that he
had difficulty getting in and out of the hole, and it looks as if it will be easier to dig
it by hand. Still, we got a fair amount out. Also removed
the Carpobrotus as planned, and lay
down the granite sand paths:
The Carpobrotus all came from a single cutting that Yvonne got a couple of years ago; we have already got rid of a large quantity, and this part filled
up two compost bays:
Moved the rosemary bushes round to where the Carpobrotus used to be; several branches of the
larger one had broken off, but they had roots on them, so planted them anyway. We'll see
how they get on. There's still plenty more work to do, but we're finally getting there.
The next step is to plant all those plants that have been cluttering up the verandah.
More discussion on the difference
between Asia
and Asian. The two terms look like they
should be the same, but the usage of the latter is described in some detail in the Wikipedia
pages Orient
(for popular uses)
and Asian
people (for official uses). Those pages probably need some tidying up, but I can't be
bothered at the moment. They basically confirm what I have observed: in Australia
“Asian” tends to mean “East Asian”, but officially it relates to the
entire continent.
It's also interesting to note that some neutral terms in common use here are considered
offensive in other countries, notably the USA. Why are people so sensitive in the USA?
Does it have something to do with the history of the country?
So now I have my shiny new 1080p projector, though my favourite channels—SBS
“HD” and ABC “HDTV”—only broadcast 720p. Still, one can hope.
But things are changing: to celebrate the arrival of digital-only TV in some parts of
Australia, ABC has dropped its HDTV programme altogether; it's now only available in 576i.
They're using the 720p signal to broadcast—continuous news! Have they gone mad? Or
is this some political power play? I'm astounded.
The Bureau of Meteorology has forecast several
frosts this month, but today was the first we experienced. Nothing significant, and the
weather station reported a lowest temperature of +0.3°. Clearly it was cooler lower down:
Now that the Carpobrotus is gone, we
can move on with planting the myriad plants we have been holding back. We have some ground
cover plants that have been waiting since last November, including two pots of this plant:
Yvonne asked if we could make more than one plant out of the
pot. Tried, and it came apart in my hands. Further examination showed a really strange
root system, which seems to develop above ground. The individual plantlets are joined by a
stem that gradually dies off. So I could chop through it and make many small individual
plants:
Spent quite some time doing that—I must have had over 50 plantlets, which I planted
round the bird bath. Even if some of them die, it should fill out pretty soon.
Also planted a little hellebore that we
bought at the Ballarat Open
Gardens last November, at the north of
the pond towards the end of the row
of buddlejas:
I had intended to plant
the Strobilanthes anisophyllus, but
we're still not in complete agreement about where it should go, and the place we had more or
less chosen proved to be full of onion grass, which I really should remove first.
I've replanted a lot of spring bulbs recently. They've been in the ground for a couple of
years, and they're expanding:
But why do they grow in a straight line? And how can you see which direction it's going to
be when you plant a single bulb?
We've been meaning to use up various odds and ends for some time, and came to the conclusion
that tacos would be the way to do it. Then
I discovered that I didn't really mean tacos,
but burritos. Scraped around to find a
recipe, and discovered in the process that we had the wrong kind of meat. Faked yet another
filling, which didn't taste too bad, but I
don't think I'll repeat it in this form.
Three years ago tonight we spent the first night in our
new house in Dereel. In the past I've referred to
9 July 2007 as the date, but we didn't actually move anything into
the house on that date. Sadly, beyond noting the date, there's nothing much else to say.
Three years used to be significant, because until I was over 30, I hadn't lived anywhere for
that long. I stayed in many places for almost three years, but didn't make it.
One of my saved searches on eBay is
for saffron. Saffron is the most
expensive foodstuff on earth—depending on where you buy it, a kilogram can cost
between $350 and $110,000. I tend to buy at the lower end of the price scale, obviously.
Today I saw once again an advertisement for 20g saffron for only $9, or $450 a kilogram.
But it said “SAFFRON FLOWER (Safflower) PURE SPICE”, so off to find out
what safflower is. It wasn't saffron,
of course, but surprise, surprise: it's a plant that I pulled out as a weed 18 months ago. Here the picture from Wikipedia, then my
own photo:
Nasty weather this morning: dark and windy, about the worst possible for the exterior photos
I do every weekend (normally Saturday), so I postponed them until tomorrow.
When chopping down the trees to the immediate north-east of the house
(Lilac, Senna), we had left one of
the Sennas, though we weren't sure that was the correct decision. The weather made its mind
up for us, uprooting the remaining Senna and breaking off a couple of cannas that hadn't
died yet:
It makes it clear that my decision to put in a wind break was a sound one. The posts are
already there; now I need to buy some cross-members to attach the screen to.
Despite the weather, into Ballarat for
the third puppy training lesson for Nemo. Didn't get
very far: it started raining, lightly at first, but then comparatively heavily. Didn't get
any photos of him either.
On the way home, turned off to look at the roadside plant shop
in Napoleons. There was nothing there,
but we saw a sign pointing to a garage sale, so on to investigate. A couple of kilometres
further on we found another sign. It had seen better times, and was now pointing either
straight ahead or to the left. This was round the place where the navigator gave us the
strange directions last week:
Which way was it? The sign said 119 Parkinsons Road, so I just needed to type it in to the navigator; and it found
it. It took us there, but the wrong way: we had just passed the junction, and it kept us on
the road we were on, taking us about 2 km longer through some of the worst roads I have been
on in the recent past. If we had turned around and gone back 20 metres, it would all have
been on good roads. Another limitation of the navigator.
There wasn't much at the garage sale, but Yvonne was in a
buying mood and bought a keyboard for her horses and a couple of toys for Nemo.
Finally got round to putting the new (Winfast
DTV Dongle Gold (who thinks up these names?) into cvr2. It worked out of the
box—that's a first—but for some reason I couldn't persuade MythTV to do a manual recording on it, possibly because of
some incorrect timing calculations. Shortly afterwards it recorded a film on it normally,
so there's nothing wrong with the tuner.
About the only issue I had wasn't directly related to the dongle: it only has an antenna in
connection, not an antenna out. One of the two existing (PCI) tuners I have in the machine
also has no antenna out, so there was no direct way of connecting all three. Over to Chris'
place to collect the third one, but it proved that she didn't have it, and I spent a fair
amount of time looking for tuners and also a display card with TV out to connect up to the
analogue TV I lent to her a while back. This is the same configuration that I set up on
18 September 2004, and which I described in some detail. At the time I
wondered whether it wasn't too much detail, but now I'm glad I did it.
Also finally got around to comparing the brightness and the size of the displays of my 3
projectors, using an xterm with white background as a basis. The results were
surprising:
|
|
Screen |
|
Screen |
|
Screen |
|
Screen |
|
Top |
|
Bottom |
|
|
|
Lumens |
|
Top |
|
Bottom |
| Projector |
|
height |
|
width |
|
diagonal |
|
area |
|
left |
|
left |
|
Centre |
|
centre |
|
right |
|
right |
|
| Acer |
|
108 cm |
|
202 cm |
|
229 cm |
|
2.18 m² |
|
58 lux |
|
70 lux |
|
106 lux |
|
231 |
|
124 lux |
|
145 lux |
| Panasonic |
|
133 cm |
|
231 cm |
|
267 cm |
|
3.07 m² |
|
148 lux |
|
151 lux |
|
185 lux |
|
568 |
|
113 lux |
|
134 lux |
| Sanyo |
|
137 cm |
|
241 cm |
|
277 cm |
|
3.30 m² |
|
147 lux |
|
174 lux |
|
241 lux |
|
792 |
|
144 lux |
|
176 lux |
|
I had expected the relationship between the brightness of the projectors, but not the light
falloff to the corners, which was barely perceptible. And the overall brightness is nothing
to worry about, I suspect. The projectors are supposed to have brightnesses of 1000 lumens,
1000 lumens, and 1200 lumens respectively, but that's before processing.
Chris around for dinner, after which we let Nemo play
with his new toys. One looks like a cross between
a Rosella and some kind of dinosaur:
It makes a groaning noise which scared Nemo off at first, but now he loves it. The question
is how long it'll last; but for $3 for two toys, I think we can survive.
The Bureau of Meteorology had forecast rain
today, but it was a nice sunny day, and my biggest problem was taking my panoramas into the
sun:
This was also the first time I tried out the dual head panorama levelling system with the
panorama head mounted on top of a ball head:
The ball head allows me to position the panorama head in a horizontal plane. Previously I
had been doing that by adjusting the legs of the tripod, which was a real pain, since the
legs aren't perpendicular to each other. The “correct” way to do this seems to
be to buy a leveling plate such as the Manfrotto 338 3416, which sells for US $95:
I'm not even convinced that the levelling plate does it better. It's limited to 5°
adjustment in each direction, and it has the same disadvantage as levelling the tripod legs,
in that the adjustments aren't orthogonal. It does have the distinct advantage of
stability; there's quite a moment on the ball head, and I have to tighten it strongly to
stop it tipping. And while adjusting it I have to take up the moment myself. But the ball
head was just lying around, and even if I had bought one, it would be considerably cheaper
than the Manfrotto plate. As it was, my total investment to date has been 32¢.
The good weather also gave us some time in the garden, which sorely needs it. Spent an hour
or so removing weeds, but didn't get as far as planting anything new.
Looking at ABC's cancellation of the HDTV
programme, there's more than meets the eye. Today I recorded the only film in
“HDTV” (in this case 1080i) on any channel during the whole week. It wasn't
HDTV:
About the only thing that's high definition is the WIN logo. The film has clearly been
scaled up. Looking through the programme, it seems that the only things that are in HD any
more are news programs, so ABC's decision may not be as stupid as it seems. But why are
they dropping HDTV content? Does it cost them more?
My discussion with SkyMesh is going
nowhere. My support requests go unanswered for up to 2 weeks. They claim that my modem is
defective because they can't reach it from time to time and ignore the fact that these
issues have occurred with three different modems. At the same time, they want me to
pay up front for a very expensive replacement. When I asked why, I got the response
(original formatting):
. We're not prepared to pay Skybridge to send you a modem because we
have no way to get it back from you if you decide to keep it
This is tantamount to accusing me of intending to steal the modem. The facts of the matter
look very different to me. Since moving to SkyMesh I have had 125 outages, measured as when
both the other end of the link and all 5 systems that I ping are unreachable. I add this
latter check because it seems that sometimes the other end of the link doesn't respond. But
I've been seeing a lot of this kind of thing:
1274988765 1 5 # Fri May 28 05:32:45 EST 2010
1274988881 1 0 freefall w3 www.auug.org.au ozlabs.org ftp.netbsd.org # Fri May 28 05:34:41 EST 2010
1274988928 0 2 w3 www.auug.org.au ftp.netbsd.org # Fri May 28 05:35:28 EST 2010
1274988955 0 4 ozlabs.org # Fri May 28 05:35:55 EST 2010
1274988992 0 3 www.auug.org.au ozlabs.org # Fri May 28 05:36:32 EST 2010
1274989028 0 3 freefall www.auug.org.au # Fri May 28 05:37:08 EST 2010
1274989054 1 3 w3 ozlabs.org # Fri May 28 05:37:34 EST 2010
1274989081 1 3 freefall w3 # Fri May 28 05:38:01 EST 2010
The interesting columns here are the second (ability to ping the other end of the link) and
five systems spread around the world. In this example, the link appears to be down, but I'm
able to ping random subsets of the five systems. This looks very much like problems in
SkyMesh's network to me. Wrote back and asked them to fix that first, and then we could see
what was wrong with the modem. I'm losing hope of an amicable settlement.
w_scan is a utility to
perform a frequency scan on DVB-T tuners.
I've tried it a couple of times in the past,
but it didn't work: it scanned only the frequencies stored in a pre-defined table, and they
weren't the Australian ones.
That seems to have changed: it now takes a parameter to specify the country, so I tried it
out again. The result: it works some of the time. It doesn't find anything at all with the
new Winfast DTV Dongle Gold, and with one of the other tuners it didn't find SBS:
634500: (time: 03:14) (time: 03:16) signal ok:
QAM_AUTO f = 634500 kHz I999B7C999D999T999G999Y999
Info: NIT(actual) filter timeout
634625: skipped (already known transponder)
...
tune to: QAM_AUTO f = 634500 kHz I999B7C999D999T999G999Y999
(time: 06:38) Info: SDT(actual) filter timeout
Info: NIT(actual) filter timeout
The problem here is that SBS broadcasts on 634.625 MHz and not 634.500. w_scan sees
the real frequency, but drops it. The filter timeout seems to be part of the issue.
A little more work in the garden. In the morning the weather was mild and there was almost
no wind—ideal weather for spraying weeds—but Yvonne was away training horses. By the time she got back, the wind had increased too much for
us to be able to do any spraying, so I spent an inordinate amount of time pulling out
(other) weeds by hand. We really must keep this under control.
As planned yesterday, I tried some other simulcast recordings which I thought might be
HDTV. This was on “WIN” (another
Microsoft related company, it seems). Here are the results from two different shows, first
the standard 567i programme, then the 1080i one:
If any of this is HDTV, they're hiding it well. I'll have to guess that this isn't the case
here. But one thing it does show is that the higher “resolution” makes the
jaggies from the interlacing less painful.
It seems that news programmes are the ones most likely to have HD content. Recorded
7 HD News (a Yahoo! company) on the SD and HD channels; the programme
guide had flagged the latter as “HD”. Result? Both in SD:
=== grog@cvr2 (/dev/pts/2) /recordings 33 -> mpid
2006_20100713175700.mpg: Seven-News 2010-07-13-1757
VIDEO: MPEG2 720x576 (aspect 3) 25.000 fps 10000.0 kbps (1250.0 kbyte/s)
2060_20100713175700.mpg: Seven-News 2010-07-13-1757
VIDEO: MPEG2 720x576 (aspect 3) 25.000 fps 10000.0 kbps (1250.0 kbyte/s)
Finally I watched a programme that I had recorded from ABC 1. The results are terrible: the
combination of deinterlacing and upscaling makes the text almost illegible. Here are two
consecutive frames, followed by a detail of the first one:
This quality can't be explained by the low resolution alone; the example above from
“WIN” is much better. I'm left with the impression that ABC is really trying to
make a political point. How can anybody get away with such poor quality? To think that I
bought a 1080p projector for this junk!
I've been half-heartedly looking for a universal and simple method of transcoding video
images for some time now. There are plenty out there, as long as you make some concessions.
Then I need to put them on my photo pages. But somehow I just
haven't got round to doing it, and Yvonne has been asking me
to do so for some time. Well, she's been asking me to put them on Youtube, something that goes very much against the
grain.
Today I couldn't procrastinate any more, and uploaded two clips that she had taken of Nemo playing with his sighing bird toy on Saturday. It started off promisingly: I went to http://upload.youtube.com/my_videos_upload and found:
That looked like some Javascript problem, like firefox frequently develops. Normally I can get it back to life by
killing firefox and restarting it, but that didn't work here. Still, I was able to
select the link, and was presented with one of these horrible file selection windows,
navigating directories one level at a time, and then for no apparent reason (literally) I
got an error message:
No option to retry; I had to go back to the horrible file selection window and try again.
And it failed again; possibly it's the Javascript issue.
Then I noticed the line at the bottom:
Upload problems? Try the basic uploader (works on older computers and web browsers).
Selected that, and was able to upload. It even has the advantage that you don't have this
horrible file selection window. Next time I'll go there directly, even if I fix this
Javascript issue.
The most obvious damage from the winds over the weekend was the uprooted Senna, but I've been discovering more. The verandah
panorama taken on 11 July 2010 shows it.
The hibiscus and one of the grasses were
nearly pulled out of their pots, and they needed stakes to hold them up:
High time to put the wind breaks in place. But beyond (finally) harvesting the remaining
potatoes, I didn't get much done. I need to stop messing around and get all this work done.
What's stopping me?
Baked not one, but two loaves of bread today: my standard
rye bread and a white bread with plain white bread flour. Last time I baked, I had a
small accident: instead of putting in a total of 860 ml of water, I put in 920. On that
occasion I decided to leave it that way and see what would happen. The results were
surprising: the bread rose more—something that I've been trying to achieve—and
seemed generally better. So today I tried even more: instead of 1300 g flour, I used 1250,
along with the same amount of water.
It would be nice to say that it was even better, but it wasn't: I forgot to change the
temperature in the oven after putting the bread in, and the surface was slightly scorched.
But I think that the result might otherwise be even better. More experiments needed, but
I'm surprised how uncritical the proportions are.
Nemo is growing up rapidly, getting stronger and more
adventurous. He's also quite a handful, and he's gradually getting on my nerves—even
on Yvonne's. How do we get him to slow down, to stop
mouthing (play-biting), to leave the cats alone? Ian has given us suggestions, but if
they're working, they're taking their time. It's also interesting how, after millennia of
domestication, people still have such radically different opinions about how to train dogs.
Yvonne is currently reading a book by the monks of New Skete, and they suggest
“massage”, apparently just stroking the dog. The intention appears to be to
draw his attention away from the undesirable activity. We'll try that for a while;
hopefully it will work.
The kangaroos have been back in force lately. I've seen up to 20 on occasion, and
Yvonne claims to have seen up to 40. I had to chase them
away twice today, and though I haven't seen any in the garden for many months, they've
clearly been there:
We have planted a lot of what I think are Chasmanthe floribunda. At one point I had thought that they were very similar to the
Watsonias, but they bloom in late summer, and
these bloom in early spring.
I'll keep an eye on them as they bloom; it looks like we'll have plenty around the garden.
This week ALDI had a netbook on special, only
$389, with a real keyboard, 250 GB disk and 1 GB memory, and the incongruous name Medion
Λκουλ (which friends claim is really something written
with fonts that imitate Greek). ALDI hardware has the wonderful advantage that if you don't
like it, you can return it for a refund within 2 months, clearly a great advantage
considering the hardware compatibility issues that free operating systems have with small
laptops.
Yvonne brought back one, and I spent the afternoon playing
around with it. It doesn't have a DVD drive, though it comes with a number of what I think
are DVDs, one of which has the interesting statement (in German) “You are not
authorized to make unauthorized copies of this data medium”.
So I need a USB DVD drive. Fortunately I bought a new drive last week, and I have an old
USB adaptor housing. Put the drive in and tried things out; worked. But it wasn't the disk
I wanted, so I still had to burn one. Put that in swamp, my test machine, and had
lots of difficulty, finally coming to the conclusion that Yet Another Motherboard has a
defective USB. I couldn't mount my external USB disk drive there (it stopped the machine
from even finishing POST), and
the USB keyboard had trouble when anything else was connected. Damn. That's the last of
that series of motherboards, so now I have plenty of perfectly good left-over memory and
processors. I wonder if I can find old motherboards on eBay at a reasonable price.
The next attempt was with a USB stick, which I burnt in teevee. It wouldn't boot; it
looks as if the image was faulty. But I did locate a DVD with FreeBSD 8.0-RELEASE, and booted from that. The first
thing I wanted to do was to back up the disk, which had to go over the network, so went into
the Fixit mode. The kernel found the (wired) Ethernet interface, but not the wireless one.
And after a lot of investigation, it proved that I couldn't receive any data via the
interface. netstat showed that the data was arriving, but it didn't seem to make it
out of the driver.
In the end put the disk from swamp (9-CURRENT from some time in March) into the USB
housing and booted the netbook from that. That worked, and so did the wired interface;
still no wireless interface, but I'll look at that problem later. Copied the disk image
to dereel, compressing it from 232 (“250”) GB to 17 GB in the process,
and showing that these Atom CPUs
really aren't very fast. Tomorrow I'll install the image on the internal disk and
investigate getting the wireless interface going.
As planned, copied the external disk to the internal disk on the
Λκουλ netbook. In the process, discovered a complete
object tree for the last OS build, clearly nothing I needed to copy across, so during the
copying (with bsdtar), removed the remaining parts of the /usr/obj tree.
Surprise: unlike gnutar, bsdtar can't handle that. It died on me without
completing the remainder of the copy. That looks like a serious bug to me: sure, people
don't normally remove files while they're being copied, but it can happen. Another reason
to stick to gnutar (the other is that I'm more used to the baroque command line
options).
After that, as planned, brought the system (FreeBSD CURRENT of about March of this year) up to date. That took forever:
>>> World build completed on Fri Jul 16 14:59:37 EST 2010
--------------------------------------------------------------
16386.13 real 24823.34 user 4963.75 sys
That's over 4½ hours elapsed time, nearly 7 hours user time and 83 minutes system time: over
8 hours CPU time in total. It would be nice to blame that on the WITNESS option
that is set by default in -CURRENT, but that should only affect the system time. The times
for the kernel build were corresponding, and even the installworld target (mainly
copying) seems to have taken longer than usual, possibly because of the low-power disk:
3759.22 real 4835.84 user 956.24 sys
299.26 real 81.68 user 121.64 sys
This is all without a valid comparison, of course, but next time I build on a
“normal” computer, I'll be able to do that comparison. More to the point, I
still don't have a wireless interface, and I can't start X:
(EE) VESA(0): No matching modes
(EE) Screen(s) found, but none have a usable configuration.
That's not surprising, considering the age of the X server, but I think I'll try Linux first
before I upgrade. If I can't get the wireless interface working, it won't help anyway.
What is that flower that I photographed yesterday? It was one of my mystery plants, but then I found out what it was and
moved it from that page. Spent a while searching for it (as I mentioned in yesterday's
entry, it's chasmanthe floribunda), but it occurred to me that I hadn't done anything sensible by removing the entries from
the mystery plants page, so spent some time reinstating them. One of the issues was that I
had changed the names of the pages (for example, http://www.lemis.com/grog/Gardening/mystery-12.php became http://www.lemis.com/grog/Gardening/chasmanthe-floribunda.php), and it's
possible that some references remain to the “mystery” name. Sure, I can put in
redirects, but how about a single file that contains the lot? In the end I wrote a page
that looks like this:
<?php
include "header.php";
$renames = <<< EOS
...
/grog/Gardening/mystery-7.php /grog/Gardening/oxalis.php
/grog/Gardening/mystery-9.php /grog/Gardening/erysimum.php
/grog/Gardening/mystery-10.php /grog/Gardening/tree-dahlia.php
/grog/Gardening/mystery-12.php /grog/Gardening/chasmanthe-floribunda.php
/grog/Gardening/mystery-13.php /grog/Gardening/chlorophytum.php
EOS;
redirectme ($renames);
?>
The function redirectme looks for the name of the page in the first column and
redirects to the page in the second column. All I needed to do was create a link to this
page with the name in the first column. No extra files (though I did use a symlink so that
I can update the page easily).
Nemo is getting calmer, but not much. From Ian's point
of view he's doing well: he learns well and enjoys the exercises we give him. But he's
a pain to have in the house, and any progress we have made with stopping his mouthing
is very slow. Even Yvonne is wondering whether we shouldn't
get rid of him. But there must be a way to solve the problem.
Yana and Sundance have made it here
on their journey from Ontario to South
Australia:
At Yana's request, a steamboat,
the first we've had in a long time:
Sundance promptly fired up his computer and tried to get a DHCP lease. Problem: we don't use DHCP.
But the satellite modem has a server, so enabled it. Result: reboot and a minute off the
network—a programmed dropout:
Start time End time Duration Badness from to
(seconds)
1279268047 1279268095 48 0.016 # 16 July 2010 18:14:07 16 July 2010 18:14:55
That took 48 seconds, with an uncertainty of about 30 seconds—noticeably less than the
other dropouts. Does this mean that the others have some other reason?
Today was puppy training for Nemo in Ballarat, but I'm so discouraged that
I didn't go. Yana and Sundance did, and Yana took
some photos with her new Canon Ixus 130. Somehow it doesn't seem have any better autofocus
than other compact cameras, and as usual there's no manual focus. The result is that the
camera decided to focus on something that it could identify easily, which was almost never
Nemo:
They did some more training—all good stuff, but it doesn't address the problems we
have:
Ian had some suggestions about how to calm Nemo down, of course, but somehow it didn't sound
convincing. Jenny Judson also found out about the problem and called up Yvonne and spoke to her for an hour, coming up with some new idea
(“The Monks of Old Skete are no longer modern, and there are also problems with the
Weston method [which Ian uses]”. As I commented a couple of days ago: it's
strange that people are still not agreed about dog training after all the millennia that
dogs and humans have been associated. Anyway, Jenny sent some documents with other
suggestions, which we'll examine in the coming days. For whatever reason, Nemo behaved
himself today.
As planned, installed Ubuntu 10.4 on
Λκουλ. Why do Linux installations take so long?
Anyway, it came up and recognized the display hardware immediately, and the Ethernet
interface was also recognized. But still no joy with the wireless LAN card. Did some
googling and discovered that yes, people had had it working, but I didn't have enough time
to follow through.
About the only thing that did become apparent: this machine has a 1024×600 display,
which I knew before I bought it, of course. And they seem to be becoming more prevalent.
All the more reason to reject window managers and programs that take up the top and bottom
of the display. After removing all the bars and status displays and things, the area
remaining for a “full-screen” firefox window is only 1012×376, barely more than half the height of the
screen, and barely more than the display height of the EGA that IBM introduced
nearly 30 years ago:
So modern display software wastes 38% of the usable area on this tiny display. That's yet
another reason to reject GNOME and other
Microsoft-like display environments and use a window manager that doesn't want to be an
entire computer.
Indian food again today. While in Waterloo, Sundance participated in a
course in chapati making, so we got him
to do the chapatis. Quite interesting: his dough was much moister, and he cooked them at a
much lower temperature:
The results were good, but the lower cooking temperature meant that it took nearly half an
hour to make them all. I must investigate. It's also interesting to note the method of
pressing down on the chapatis with a cloth; I had thought that a spatula would do the job,
but it seems that the cook in Waterloo had insisted on the cloth.
Dinner also brought home to us the difference between our appetites and the appetites of
people who have been travelling by bike for 10 months. I had already expected that, but
Yana and Sundance must each have eaten 5 times as much as
Yvonne. Sundance was also talking about the preference of
Australian research institutes to perform practical experiments rather than theoretical
research, so considered dividing
a quark, with mixed results:
The last few days with Nemo haven't been much fun, but
there are signs the worst is over. We've taken heed of what a lot of people have said, have
been more gentle in the punishment (and yes, we believe in punishment), haven't tried to be
Alpha dogs (against Weston's recommendation), and other odds and ends. And hey, it seems to
be working! He's much more gentle and controlled. But what is the real reason? It would
still be nice to know.
So I have Ubuntu 10.4 running on
Λκουλ, but without the all-important WLAN interface.
Upgraded the kernel to the latest version, without any obvious improvement, and then spent
more time looking around the web and found some instructions on how to
compile the driver. Did that—another 200 odd MB of data—and once again there
was no improvement.
OK, the reason I tried Ubuntu is that it might have been easy. I always find so many things
that grate on Linux, and for some reason that I didn't care to follow, the NFS file systems
would no longer load—instead I got the startling message Disconnected from
Plymouth, which doesn't seem to have any further meaning. But the text-mode terminals
were missing. I'm not sure why, but using this tiny interface with the funny mouse to do
anything is so painful that I couldn't be bothered. Back to investigate FreeBSD.
But how? Ubuntu had not been able to install its bootstrap in the root partition, so it had
overwritten my FreeBSD boot selector. OK, I've
used GRUB before, so went off looking
for the config file. Surprise, surprise! The authors had recognized that GRUB was a pain
to use, so they rewrote it and replaced it by something that is an order of magnitude more
painful to use. All I wanted was to add an entry for the FreeBSD partition on the disk,
which should have been a couple of lines. To ensure a smooth transition, they've changed
the names of the configuration file, but finally found the official manual for GRUB 2,
conveniently numbered 1.98-r2508, and discovered that the configuration file is now
called /boot/grub/grub.cfg, which starts of like this (somewhat trimmed):
# DO NOT EDIT THIS FILE
#
# It is automatically generated by /usr/sbin/grub-mkconfig using templates
# from /etc/grub.d and settings from /etc/default/grub
if [ -s $prefix/grubenv ]; then
load_env
fi
set default="0"
if [ ${prev_saved_entry} ]; then
set saved_entry=${prev_saved_entry}
save_env saved_entry
set prev_saved_entry=
save_env prev_saved_entry
set boot_once=true
fi
...
So it's no longer a configuration file: it's a script. At the very least they could have
separated configuration information and code. Took a look in /etc/grub.d,
which, if anything, was worse. It's all just scripts.
Finally found instructions, to do exactly what I wanted: add FreeBSD to a Debian boot. What I
ended up with was a file /etc/grub.d/40_custom with the following content:
#!/bin/sh
exec tail -n +3 $0
# This file provides an easy way to add custom menu entries. Simply type the
# menu entries you want to add after this comment. Be careful not to change
# the 'exec tail' line above.
menuentry "FreeBSD" {
set root=(hd0,1,a)
chainloader +1
}
The meaning of chainloader is not really clear from the manual, which doesn't give
any BSD examples, though it does state that chainloader is not necessary for them.
I also didn't remove the comment, though I should have done, since it's completely wrong.
This isn't easy. The idea of modular configuration files is arguably useful, but mixing
code and data in a “configuration file” is most certainly not. To quote one of
the comments in the previous instructions, written by “doik”:
i install grub 2. it is foul beyond words to configure and is not pretty at all. it also
seems much slower at getting to the menu. grub 2 is FAIL.
So, finally I had a GRUB configuration that should boot FreeBSD. Rebooted. No menu popped
up, and I ended up back in Linux. Read the manual again. Played around with variables that
might potentially change the behaviour. No change. Finally, Juha Kupiainen told me that I
had to hold down the left shift key when booting. That worked: I got a menu and was able to
select FreeBSD. But there's nothing anywhere in the manual that mentions this.
Somebody told me that this is a (presumably undocumented) Ubuntu extension. What a pain!
Still no wireless card, of course. pciconf -lv tells me:
none2@pci0:3:0:0: class=0x028000 card=0x817210ec chip=0x817210ec rev=0x10 hdr=0x00
vendor = 'Realtek Semiconductor'
device = 'Realtek RTL8191SE wireless LAN 802.11N PCI-E NIC (RTL8191SE ?)'
class = network
I wonder what the question in brackets is intended to mean. More googling, and found
instructions for installing this code, first requiring downloading the drivers from
the Realtek web site, which led me around a
couple of cycles in a link loop before starting to download two copies. Left that until
tomorrow.
After dinner, had another discussion with Sundance about the current
understanding of the Standard
Model, illustrated with lots of leftovers:
To my great disappointment, I found that the universe is not all Quark: there
are leptons (on the left)
and bosons (not shown here). I think we
should change the name of leptons to Hartkäse. And of course, compared to the Wikipedia
page, the arrangement is different:
The whole thing is on its side, the bosons are missing, and the colours are roughly
reversed—giving us the right to claim that this is the non-standard model, though
Sundance has also voiced the opinion that Wikipedia has it wrong. I recalled the Standard 8
model of the late 1940s, and thought that it was somehow related to Ford Prefect, but that
proved to be another manufacturer. We're currently considering whether it would have more
success if it were called the Triumph Model.
We had intended to eat huevos
rancheros for breakfast this morning, but real life got in the way, and Yvonne had to go into town. In the meantime, Sundance and Yana prepared for the last leg of their journey
to Adelaide. They were planning to make
it to Lake Bolac tonight, and they
left shortly after midday:
Ah, what peace! It was fun having them, but basically we're happiest by ourselves.
Definitely not the unlimited social types. Spent the rest of the day relaxing.
I haven't been playing with λκουλ much lately—I'll
take a stab at the NDIS-based wireless network later—but I wanted to boot it in
FreeBSD today to try out something potentially
dangerous. And, of course, the Left shift trick didn't work: I had no way of booting
into anything but Linux.
This is with GRUB 2, and there's
supposed to be a way of getting it to present the boot menu automatically, like there was
with GRUB 1. But everything has changed. The files are spread across three directories:
global information is in /etc/default/grub, “configuration” information
(consisting largely of scripts) in /etc/grub/, and the boot environment is
in /boot/grub/, all 186 files and 4.3 MB of it. I've understood some of the puzzle,
but not how to get the menu. It looked as if it should be an entry
in /etc/default/grub, but I couldn't see one. There are comments for some of the
values, like the all-important resolution of the graphical display—for a boot
loader!—but not for the menu. This is clearly a case of FTFM: it's almost impossible
to find out how to configure the thing.
With the help of a number of people on IRC, finally got it to work. Ubuntu has its own document, apparently written as an
afterthought, and at the section with the emetic URL https://help.ubuntu.com/community/Grub2#/etc/default/grub%20(file)—with spaces in it!—I found the answer:
undefine GRUB_HIDDEN_TIMEOUT. It seems that this should have happened
automatically: one of the myriad scripts, /etc/grub.d/30_os-prober, should have
detected FreeBSD, but it didn't. So, finally, after a couple of hours, I can dual boot the
machine the way the FreeBSD boot0 boot manager would do without any configuration.
Isn't progress marvellous?
Access to http://www.lemis.com/ failed today. It's hosted at
TransACT, and further investigation showed
that I couldn't access that network from 09:54:17 until 10:33:16, nearly 40 minutes.
Traceroutes from this end ended at Equinix:
=== grog@dereel (/dev/ttypa) ~/public_html 225 -> traceroute ozlabs.org
traceroute to ozlabs.org (203.10.76.45), 64 hops max, 40 byte packets
1 sat-gw (192.168.5.100) 0.655 ms 0.330 ms 0.307 ms
2 192.168.1.33 (192.168.1.33) 779.460 ms 677.721 ms 636.202 ms
3 203.169.16.37 (203.169.16.37) 606.518 ms 808.812 ms 619.523 ms
4 203.169.16.25 (203.169.16.25) 683.201 ms 873.447 ms 621.327 ms
5 203.169.17.38 (203.169.17.38) 624.849 ms 870.203 ms 621.259 ms
6 gw2.nsw.skymesh.net.au (203.34.110.250) 618.183 ms 840.786 ms 616.479 ms
7 p7718.syd.equinix.com (202.167.228.61) 652.372 ms 883.430 ms 608.313 ms
8 * * *
I was able to access TransACT from my US server, so it wasn't TransACT itself. A
traceroute in the other direction looked like:
=== grog@bilbo (/dev/pts/2) ~ 2 -> date; traceroute sat-gw-ext.lemis.com
Tue Jul 20 10:10:59 EST 2010
traceroute to sat-gw-ext.lemis.com (180.181.112.227), 30 hops max, 60 byte packets
1 pesto.tip.net.au (203.10.76.16) 0.163 ms 0.162 ms 0.152 ms
2 ve2657.core01.gate.transact.net.au (202.55.146.85) 0.395 ms 0.474 ms 0.554 ms
3 gigabitethernet1-1.core01.eqnx.transact.net.au (202.55.144.82) 4.131 ms 4.222 ms 4.290 ms
4 * * *
Probably not SkyMesh's fault, but they're
the only people I could report it to. I'm too angry with them at the moment to call them up
and risk losing my temper, so sent a mail message at 10:23:51. No response, no
acknowledgement of receipt. I wonder to what extent the resolution of the problem was
related. Certainly that would mean that they reacted pretty quickly (and that thus the
problem was at SkyMesh after all).
The weather wasn't spectacular today, but I did manage to get some weeding done. It's
amazing how much grass has popped up in the area where we first started removing the lawn
and covering it over with cardboard and mulch over two years
ago:
Spent about half an hour, until it started raining, and only got a couple of square metres
weeded. Once this is finished, I should really keep on top of it.
dereel has been up for 110 days now. Or it was. While watching TV in the evening,
we had a power failure, this time over 3½ hours. Yes, we now have UPS running just about
everything, but they don't last that long. At least we didn't damage the new projector.
The power came back shortly after midnight, turning the heating on to daytime levels and
turning on many lights. But by morning, the hot water (storage system, heated only at
night) still hadn't come up to temperature. I wonder why: that should have been plenty of
time. Turned the switch on to the expensive daytime position, and it warmed up, so at least
the heater isn't damaged. We'll see tomorrow if something else has gone wrong.
Sometimes I think we're making progress with Nemo, and
at other times I despair. Today he ran all over the garden, wouldn't come back, and finally
saw some kangaroos and ran after them. It's a good thing he can't get through the fence up
against the lagoon.
These weeds in the bottom corner of the garden are particularly obnoxious. Spent the best
part of an hour mainly clearing another square metre or so.
And the kangaroos are attacking the plants again! It's been months, but not enough. Back
with the protection. Maybe we should put a high fence around the whole garden to keep out
not just the kangaroos, but also Nemo.
Also spent some time measuring things up in preparation for a timber purchase, maybe
tomorrow.
Continued trying to get the wireless network card running on
λκουλ today. I had downloaded the driver from the
Realtek site, and the instructions were straightforward enough. But where were the drivers? The
instructions mention the files net8192se.inf and rtl8192se.sys, but what I
got was a whole directory tree. Went finding and found four
files, RTL8191_8192_SE_WindowsDriver_2016.2.0521.2010.F0062_23.P0525_ISS_1.00.0157.Win7.L/91_92_SE_Driver/Win7X64/net8192se.inf,
RTL8191_8192_SE_WindowsDriver_2016.2.0521.2010.F0062_23.P0525_ISS_1.00.0157.Win7.L/91_92_SE_Driver/Win7X86/net8192se.inf, RTL8191_8192_SE_WindowsDriver_2016.2.0521.2010.F0062_23.P0525_ISS_1.00.0157.Win7.L/91_92_SE_Driver/Win7X64/rtl8192se.sys and
RTL8191_8192_SE_WindowsDriver_2016.2.0521.2010.F0062_23.P0525_ISS_1.00.0157.Win7.L/91_92_SE_Driver/Win7X86/rtl8192se.sys.
Presumably the Win7X64 files were the ones for 64 bit mode and the Win7X86
are for 32 bit mode (and not 86 bit mode, as the convention would seem to imply). But
what about Win7? That looks like Microsoft “Windows 7”, which is not
what I need (though it's what came with the machine). Tried running ndisgen
against them, and it happily generated a KLD that I could load. With verbose boot
messages, it also produced quite a bit of information:
pci0: driver added
found-> vendor=0x8086, dev=0x27d8, revid=0x02
domain=0, bus=0, slot=27, func=0
class=04-03-00, hdrtype=0x00, mfdev=0
cmdreg=0x0006, statreg=0x0010, cachelnsz=8 (dwords)
lattimer=0x00 (0 ns), mingnt=0x00 (0 ns), maxlat=0x00 (0 ns)
intpin=a, irq=16
powerspec 2 supports D0 D3 current D0
MSI supports 1 message, 64 bit
pci0:0:27:0: reprobing on driver added
found-> vendor=0x8086, dev=0x27da, revid=0x02
domain=0, bus=0, slot=31, func=3
class=0c-05-00, hdrtype=0x00, mfdev=0
cmdreg=0x0001, statreg=0x0280, cachelnsz=0 (dwords)
lattimer=0x00 (0 ns), mingnt=0x00 (0 ns), maxlat=0x00 (0 ns)
intpin=b, irq=19
pci0:0:31:3: reprobing on driver added
pci1: driver added
pci2: driver added
pci3: driver added
found-> vendor=0x10ec, dev=0x8172, revid=0x10
domain=0, bus=3, slot=0, func=0
class=02-80-00, hdrtype=0x00, mfdev=0
cmdreg=0x0007, statreg=0x0010, cachelnsz=8 (dwords)
lattimer=0x00 (0 ns), mingnt=0x00 (0 ns), maxlat=0x00 (0 ns)
intpin=a, irq=18
powerspec 3 supports D0 D1 D2 D3 current D0
MSI supports 1 message, 64 bit
pci0:3:0:0: reprobing on driver added
pci4: driver added
pci0: driver added
found-> vendor=0x8086, dev=0x27d8, revid=0x02
domain=0, bus=0, slot=27, func=0
class=04-03-00, hdrtype=0x00, mfdev=0
cmdreg=0x0006, statreg=0x0010, cachelnsz=8 (dwords)
lattimer=0x00 (0 ns), mingnt=0x00 (0 ns), maxlat=0x00 (0 ns)
intpin=a, irq=16
powerspec 2 supports D0 D3 current D0
MSI supports 1 message, 64 bit
pci0:0:27:0: reprobing on driver added
found-> vendor=0x8086, dev=0x27da, revid=0x02
domain=0, bus=0, slot=31, func=3
class=0c-05-00, hdrtype=0x00, mfdev=0
cmdreg=0x0001, statreg=0x0280, cachelnsz=0 (dwords)
lattimer=0x00 (0 ns), mingnt=0x00 (0 ns), maxlat=0x00 (0 ns)
intpin=b, irq=19
pci0:0:31:3: reprobing on driver added
pci1: driver added
pci2: driver added
pci3: driver added
found-> vendor=0x10ec, dev=0x8172, revid=0x10
domain=0, bus=3, slot=0, func=0
class=02-80-00, hdrtype=0x00, mfdev=0
cmdreg=0x0007, statreg=0x0010, cachelnsz=8 (dwords)
lattimer=0x00 (0 ns), mingnt=0x00 (0 ns), maxlat=0x00 (0 ns)
intpin=a, irq=18
powerspec 3 supports D0 D1 D2 D3 current D0
MSI supports 1 message, 64 bit
pci0:3:0:0: reprobing on driver added
pci4: driver added
But that was all. It should have generated probe messages for an interface ndis0,
but nothing happened. So what's wrong now? Wrong drivers? Or something else? It would
have made more sense to take the drivers that came with the machine, but they're happily
packed up in a disk image. It seems that I could use mdconfig on it if it weren't
compressed, or geom_uzip for it if it had been compressed with mkuzip. The
image is 250 GB in size, so it would be nice to be able to pipe the output of bzcat
into mkuzip, but mkuzip wants a disk file or a character device. It was easy
enough to hack it to read from stdin, but it wants to issue a call to stat
to find the size of the file. A little more work and another flag should solve that
problem. Manãna, at the earliest.
There doesn't seem to be an obvious solution to problems I'm having with SkyMesh, so today I called up the Department of
“Broadband”, Communications and the Digital Economy to ask for help. The
number they gave was 1800 254 649, but though that's the Department, it's the wrong number.
The correct number is 1800 883 488. Called that and got a recorded message telling me that
they were too busy to take my call, that I should leave my phone number and they would call
me back. I did, they didn't.
A lot of things in the garden are waiting on supplies: the pond, the wind breaks and the
vegetable garden. Finally got my act together and into town to pick up timber and pond
liner. The timber was straightforward enough, though the total cost, exactly $500.00,
looked fishy. Getting the pond foil was another matter. The cheapest was at Bunnings, not my favourite place. They sell it by
the (4 m wide) metre for $34.88 per metre. I needed 5.5 m, so that would have set me back
$191.84. Found some pre-packed foil that was much cheaper: 4.5 m, also 4 m wide, for
$115.00; from the roll it would have been $156.96. Clearly worth investigating. It's too
short, of course, but one end of the pond is only half the width, so I could cut some off
one side and glue it on the end. But how?
Found a bloke there with a “we are the experts” emblem on his name badge, who
professed not to have any idea. They did find a number of people who discussed the matter
and thought of various possibilities, but found nothing that was clearly intended for this
purpose. In the end left with a tube of swimming pool adhesive, which doesn't fill me with
confidence.
Now that the federal election has been called, we're inundated with mail, including
instructions for how to vote by post (but not when the election is), and the principal
candidates outlining reasons why I should vote for them.
At least that's the theory. Received a document
from Sarah Henderson, the
“Liberal” candidate
for Corangamite,
outlining reasons to vote for her: the incumbent party and representative are
horrible, horrible and incompetent people. She gives a list of the blunders they
have made, including “walking away” from planned measures against climate
change—the result of successful opposition by her party to anything threatening the
status quo. And she has grown up in Geelong. And that's all there is in the letter.
There is an accompanying brochure describing what she plans to do: oppose Labor and the
Greens, install CCTV cameras in public areas, widen
the Princes Highway, but nothing
much about climate change. Clearly her highest priority is to bash her political opponents.
From the content on her web site a
couple of days ago (original markup):
Today, 20th July, I am on the campaign trail in Belmont and Ocean Grove. This
morning, I was set to debate Darren Cheeseman on ABC Radio Ballarat but at the last
minute
he pulled out. As I told WIN-TV, Mr Cheeseman is continuing to hide behind
his
deceptive media releases and electoral advertisements. And despite Labor’s
deceptive and dishonest claims, Workchoices is dead and buried. But as Kevin Rudd
himself experienced, Labor will stop at nothing.
There is other stuff there too, but this is first paragraph and the only one highlighted in
<strong>. Clearly it's intended to be the most important
part.
That's not the first time that I've seen such stupidity from the Liberals. Last year I received mail
from Senator Michael
Rolandson promising not to even the budget for 30 years. They seem to have changed
their tune on this now, and claim to be able to carry it off in 3 years, now that they want
to be reelected. How much credence should you give to somebody who changes his tune that
much when they want to have some gain from it?
A clear reason to vote for Labor, one might think. But
they're almost as bad. About the only important difference on Darren Cheeseman's site is that it is
considerably out of date. The last entry is dated 11 June. But the tone is the same, and
the spelling and markup here are original:
Phoney Tonys party walks away from the Princes Highway
Wednesday, 26 May 2010 09:20
The Federal Opposition has abandoned its promise to duplicate the Princes Highway between
Winchelsea and Colac, exposing the spin and dishonesty of Phoney Tony's candidate for
Corangamite Sarah Henderson.
Read more...
Interestingly, this information is completely out of date: the Liberals appear to have
secured the promise, perhaps because of these polemics. About the only difference I can see
is that Henderson seems to consider the Internet a useful means for disseminating
information, while Cheeseman appears to have no idea. The markup of his web site looks
horrible, though it is almost standards conformant, better than Henderson's. Of further interest is that both web sites are running Apache under
Linux.
I have the distinct impression that the major parties have forgotten the purpose of an
election and consider the voting populace stupid. Maybe they are on average, but it's still
insulting treatment.
The Australian electoral
system requires a numerical order of preference for every candidate, so I can't just
omit Labor and the Liberals. Instead, I will put them in the last two places; I'll decide
later on the order. One of the two candidates will still win, of course, but it's not clear
that it will be the one with the most primary votes (votes where they are numbered 1 in the
list). In the last election, Stewart McArthur (Liberal) got 44.70% of the primary vote, and Darren Cheeseman (Labor) got 41.91%
of the primary vote, but after preferences it was Cheeseman who ultimately won, by a
comfortable 1.7%. If more people vote for the alternative parties, it might shake the big
two into rethinking the way they treat their constituents.
Watching Désaccord parfait (called “Twice upon a time” in English) on TV in the evening. It's much better
than the English ratings suggest. Came across this image:
I asked Yvonne “What do you call a horned
tiger?”. Her immediate response: “cocu”.
On one of our plant-buying sprees
in Napoleons we bought a small
succulent in ceramic pot, clearly intended for indoors. It looked a little scruffy round
the edges, but the price was right and it showed promise. We put it on the mantelpiece in
the dining room, and to my surprise, it's flowering—in the middle of winter:
That's not the only one.
Our Crassula falcata—also
somewhat scruffy, despite the high price we paid for it—finished blooming months ago
(first photo), but it is also coming up with what must be the start of new growth, though
the pink colour is somewhat unusual:
To be fair to Darren Cheeseman, I received a personalized letter from him today (“Dear Gregory”; I wonder if
he would appreciate me calling him Darren), outlining what he wants to do for his
constituency if he is re-elected, in particular the matter of health care reform. It would
be nice to be able to say that he didn't slate the Opposition, but there was a whole page in
the accompanying brochure with headlines like
“Mr Abbott's record of failure
as Health Minister” and “Mr Abbott can't be trusted on health”. But in
comparison to Sarah Henderson it
was pretty low key.
But what do I want from the Government? Technological infrastructure, mainly, and that
doesn't seem to be an issue for any candidate. Where's the world-class Internet connection?
The moderately reliable electricity grid? The high definition television? None of these
seem to interest anybody very much, not even this stupid Internet censorship scheme. Modern
technology is still far from being mainstream.
More playing around with the WLAN issues on λκουλ today.
Spent some time discussing where the firmware was; it's not in the distribution I got from
Realtek, and the instructions that I had didn't mention any, but the documentation does mentions
firmware. Callum Gibson found some somewhere, but it wasn't clear if it matched or not. He
also found a Microsoft “Windows XP” driver for the chip set, and Sue Blake sent
me the drivers she had on her identical Λκουλ netbook.
Tried Callum's first, without success—until I realised that kldload
rtl8192se_sys does not respect the current working directory; instead it uses the value
of the sysctl kern.module_path unless a directory component is specified.
Finally got the correct module loaded, and—success!
ndis0: <Realtek RTL8191SE Wireless LAN 802.11n PCI-E NIC> port 0xe800-0xe8ff mem 0xfebfc000-0xfebfffff irq 18 at device 0.0 on pci3
ndis0: [MPSAFE]
ndis0: [ITHREAD]
ndis0: NDIS API version: 5.1
ndis0: 11b rates: 1Mbps 2Mbps 5.5Mbps 11Mbps
ndis0: 11g rates: 24Mbps 36Mbps 48Mbps 54Mbps
pci4: driver added
And I had an interface:
=== root@swamp (/dev/pts/0) /src/FreeBSD/NDIS/Realtek-8191/realtek-XP/guts 110 -> ifconfig ndis0
ndis0: flags=8802<BROADCAST,SIMPLEX,MULTICAST> metric 0 mtu 2290
ether 1c:4b:d6:83:dd:a6
media: IEEE 802.11 Wireless Ethernet autoselect (autoselect)
status: no carrier
No carrier? Tried bringing it up, but it seems that that's not the way you do it any more.
Instead you need to create the real interface:
=== root@swamp (/dev/pts/0) /src/FreeBSD/NDIS/Realtek-8191/realtek-XP/guts 113 -> ifconfig wlan0 create wlandev ndis0
=== root@swamp (/dev/pts/0) /src/FreeBSD/NDIS/Realtek-8191/realtek-XP/guts 114 -> ifconfig wlan0
wlan0: flags=8802<BROADCAST,SIMPLEX,MULTICAST> metric 0 mtu 1500
ether 1c:4b:d6:83:dd:a6
media: IEEE 802.11 Wireless Ethernet autoselect
ssid "" channel 1 (2412 MHz 11g)
ifconfig: missing or corrupted regdomain database
What does that message mean? Went looking, and it seems that there should be a
file /etc/regdomain.xml with information about regulatory definitions. And I had
none. It was easy enough to find one, but why wasn't it there in the first place? My guess
is that it's because of a glitch in the update process: this disk was cloned from my test
box, as the name swamp shows. After that I got a better result:
wlan0: flags=8802<BROADCAST,SIMPLEX,MULTICAST> metric 0 mtu 1500
ether 1c:4b:d6:83:dd:a6
media: IEEE 802.11 Wireless Ethernet autoselect
ssid "" channel 1 (2412 MHz 11g)
country US authmode OPEN privacy OFF txpower 0 bmiss 7 mcastrate 0.5
mgmtrate 0.5 scanvalid 60 protmode CTS bintval 0
mcastrate appears an unfortunate choice of keyword name. But then I tried to use
it, and got a subtly different message:
=== root@swamp (/dev/pts/0) /src/FreeBSD/NDIS/Realtek-8191/realtek-XP/guts 129 -> dhclient wlan0
DHCPDISCOVER on wlan0 to 255.255.255.255 port 67 interval 7
DHCPDISCOVER on wlan0 to 255.255.255.255 port 67 interval 14
send_packet: Network is down
DHCPDISCOVER on wlan0 to 255.255.255.255 port 67 interval 21
send_packet: Network is down
^C
=== root@swamp (/dev/pts/0) /src/FreeBSD/NDIS/Realtek-8191/realtek-XP/guts 130 -> ifconfig
wlan0: flags=8c43<UP,BROADCAST,RUNNING,OACTIVE,SIMPLEX,MULTICAST> metric 0 mtu 1500
ether 1c:4b:d6:83:dd:a6
inet 0.0.0.0 netmask 0xff000000 broadcast 255.255.255.255
media: IEEE 802.11 Wireless Ethernet autoselect (autoselect)
status: no carrier
ssid "" channel 1 (2412 MHz 11g)
country US authmode OPEN privacy OFF txpower 0 bmiss 7 mcastrate 0.5
mgmtrate 0.5 scanvalid 60 protmode CTS bintval 0
Why that? I still don't know. How do you even investigate this kind of problem? The
access point is working, because Yana and Sundance used it last
weekend. It supports 802.11g, so that's not the issue. Probably the next message to look
at is one I've seen in the dmesg output:
RF is in progress, need to wait until rf chang is done.
But that doesn't happen every time I create and destroy the interface, so it could be a red
herring.
I really don't have much in the way of an excuse not to continue working in the garden, but
somehow that didn't help much. A little more weeding—I despair of getting this corner
of the garden weed-free—and some playing around setting up the wind breaks.
Yvonne went to town for Nemo's last puppy lesson, and came back via a Farmer's Market, bringing some plants
with her, a ground cover Stachys
byzantina, also called “Lamb's Ears”:
I'm still having difficulty getting the WLAN running on
λκουλ. Somebody suggested that I needed to
configure WPA, something
that I don't use, but they thought it might need it anyway. So ended up with this
in /etc/wpa_supplicant.conf (who thinks of these terms?):
network={
priority=0
ssid="allmine"
scan_ssid=0
key_mgmt=NONE
}
Added the following to /etc/rc.conf:
wlans_ndis0="wlan0"
ifconfig_wlan0="WPA DHCP"
Not that I use DHCP either, but that was the suggestion. And the result? Nothing, not
even an ndis0 interface. I had to kldunload and kldload it to even
get that. And wpa_cli couldn't talk to the WPA supplicant (who thinks of these
terms?). Nothing I have been able to do so far has helped. I suppose I should look at the
source of the Linux driver. But that doesn't work either since some recent kernel
changes. I wonder why I should bother; maybe I should just return the machine.
Leg of lamb for dinner tonight. I have to confess: I don't always make gravy the way I
should; instead I use some pre-prepared stuff called Green's Gravy Granules. It's not perfect, but with a
little bit of tweaking it makes a reasonable and quick gravy.
But how much? I've complained about this before. On that occasion I couldn't find my measuring spoons. Today I found them—bought
in Australia but apparently corresponding to the US standard—and measured 5 tablespoons. Of course, the weight of a level
tablespoon depends on how tightly you pack it. Here a full tablespoon and one that is only
half full—with exactly the same content, the second time after pressing it down:
The Australian “tablespoon” is 20 ml, over 35% bigger than the US
“tablespoon”, though I had forgotten that detail when measuring things. Last
time I got 50 g from 100 ml of granules; this time I should have got 37 g, but in fact I got
over 41 g:
Last time I had decided that 30 g per 500 ml would be about right; in fact, this quantity
proved to be about right. This time, though, I also found more information: according to
the label, the jar contains 120 g and provides 16 “servings” of ¼ “cups” (60 ml, which would make the
“cup” 240 ml instead of the defined 250 ml). That means 60 g for 500 ml, 50%
more than I find acceptable. But it doesn't agree with the weights of the
“tablespoons”. Why do people get themselves into such a mess?
Into town this morning to Victoria Park to visit Carol of the Ballaarat Dog Obedience Club (no typo) to
talk about tracking training. I'm not sure why: the trackers had all gone to Haddon for
some trials, and when we got there, she was just about to start some scheduled obedience
classes, and she had no time for us. It's quite an impressive lineup they have there, much
more than the Happy Dog people:
Watched for a while, and then off to Haddon to look for the Rifle Range, significantly
hampered by the fact that I had left my navigator at home. Finally found the place—we
had driven straight past it the first time—and found that basically it consisted of
individual dogs with their handlers and a judge who disappeared into the forest. There
wasn't really much to see, so back home again.
I can't think of anything else that I can easily do with
λκουλ to get the WLAN to work. I'll put Microsoft back
on it and see if I can get it to work like that. It's beginning to look as if it'll go
back.
Started backing it up, which first involves creating an enormous file full of zeroes to make
the disk image more compressible. On the FreeBSD partition it was about 180 GB, and it took a surprising amount of time to delete. Given
my previous problems with deletion under Linux, timed it there. Surprise!
=== grog@lkoyl (/dev/pts/0) /var/tmp 10 -> ls -l foo; time rm foo
-rw-r--r-- 1 grog grog 87129030656 2010-07-25 15:08 foo
real 0m0.266s
user 0m0.000s
sys 0m0.256s
That's exactly the opposite of what I've been experiencing. I should have gone back to
FreeBSD and measured things there, but it takes too long. Certainly it's an indication that
there's more to this issue than I thought.
It seems that we only ever do work in the garden when the weather is pleasant. That would
certainly explain why I have done so little in recent weeks. Today was a nice sunny day,
and I did a reasonable amount of work. Did some digging in the pond, hampered more by
considerations of what it should look like than by the effort. The “soil” here
is almost pure fine sand, and it's very easy to dig and to form. Also called up Bostik and confirmed that the adhesive I bought for the
pond liner would work—I think: the person on the phone seemed to think that I was
going to glue it to something else, and I had difficulty convincing her otherwise. The
other possibility, she said, was “Multibond”, which I didn't investigate.
Also mowed the lawn in the ex-cathedral, pruned some of
the Buddlejas, and finally planted
the Itea ilicifolia, which has been
looking quite unhappy:
I fully expected it to be root-bound, but in fact the roots were only about half way down
the pot, and somehow it was waterlogged: unexpectedly, it seems that the soil I used was not
draining well enough. It's in the ground now, hardly recognizable; hopefully it'll survive:
Started copying the Microsoft image to λκουλ today.
=== root@swamp (/dev/pts/0) /var/spool 3 -> bzcat /src/Images/LKOYL.bz2 > /dev/ad0
Then it occurred to me that that might not be optimal for the disk. Ran iostat and
found:
=== root@swamp (/dev/pts/1) ~ 4 -> iostat 1
tty ad0 da0 pass0 cpu
tin tout KB/t tps MB/s KB/t tps MB/s KB/t tps MB/s us ni sy in id
0 79 4.00 5153 20.13 0.00 0 0.00 0.00 0 0.00 14 0 22 4 60
0 79 4.00 5091 19.89 0.00 0 0.00 0.00 0 0.00 9 0 20 5 66
0 79 4.00 5165 20.18 0.00 0 0.00 0.00 0 0.00 14 0 18 4 65
0 79 4.00 5072 19.81 0.00 0 0.00 0.00 0 0.00 11 0 22 5 62
5000 transfers a second! Clearly this was into a big on-disk cache, but I was surprised
that the interface could take that many transfers.
Decided that it would be better to block the writes into, say, 32 kB transfers.
=== root@swamp (/dev/pts/0) /var/spool 5 -> bzcat /src/Images/LKOYL.bz2 | dd ibs=8 obs=32k of=/dev/ad0
=== root@swamp (/dev/pts/1) ~ 6 -> iostat 1
tty ad0 da0 pass0 cpu
tin tout KB/t tps MB/s KB/t tps MB/s KB/t tps MB/s us ni sy in id
1 81 32.00 53 1.65 0.00 0 0.00 0.00 0 0.00 11 0 26 0 63
0 78 32.00 82 2.56 0.00 0 0.00 0.00 0 0.00 26 0 39 0 35
0 78 32.00 84 2.62 0.00 0 0.00 0.00 0 0.00 22 0 37 0 40
0 78 32.00 82 2.56 0.00 0 0.00 0.00 0 0.00 24 0 42 0 34
0 78 32.00 86 2.68 0.00 0 0.00 0.00 0 0.00 18 0 40 0 42
Amazing! Why should it slow down at all, let alone by so much?
Finally finished copying the image and fired it up for the first time ever. It was also the
first time I used “Windows” 7. They've made the EULA a whole lot shorter now:
The first boot took for ever; it seems that it does all its adaptation to the local hardware
after delivery, and it must have taken about 15 minutes. When it finally said it was ready,
the wireless connection didn't work—if I asked ipconfig. According to the
hardware device window, it was working. Finally found a control windows similar to what
Microsoft XP uses (look at the minuscule icon bottom right, which some say looks like a
radiating antenna dish), and played around with it. It's not clear what was wrong, but it
was quite happy to take my settings, and didn't even want to reboot afterwards.
Got hold of a firefox, but
there seems to be no way to get Emacs bindings under Microsoft. I'm not sure it makes much
difference: this tiny display is too small for anything useful. I might as well take it
back. That'll also save me further pain with the WLAN connection.
Still, one thing is amusing, the name λκουλ. I got a
message from Damon Blum with the subject line “Damon Blom akoua ports
net/bwi-firmware-kmod Greg Lehey ”. So he recognized that the “Λ”
was a broken “a”, but not that the “Υ” was a
“Y”. Andy S (he keeps his surname secret) also opined that you can't do a
Google search on “ΛΚΟΥΛ”. But you can, and they're not even all my hits.
Two are mine, two contain references to my pages in the hit text, but not in the page
content, and two are books in http://books.google.com/: the
epistle to the Hebrews, 12:4, and Xenophōntos kyrou paideias biblia oktō . The characters recognized are nothing like
“λκουλ”. The first is clearly the text
“ἀκὀμι”, and the other is “δοκοῦσιν”. That's remarkably inaccurate
recognition:
Another mild frost today, less than usual, only about 1 mm of ice on the bird bath:
It's been 18 months since the Optus people
came here to look at sites for their phone tower. Since then I've heard from Greg that the
anti mobile tower fanatics have killed the project. They're putting in a tower
at Corindhap, but as he says, the
location and the lie of the land is such that we won't get much benefit. Not much I could
do about that.
But they still haven't given up. Today I found another flyer in the letter box, a printout
of a site http://savedereel.com/. More nonsense, and once again I
felt obliged to answer. In the process discovered
another web site, http://dereel.com.au/, which claims to be “A
local site for local people”. It's relatively new, and currently there's only one
topic of discussion: the mobile phone tower. A poll on that site shows that 67% of
respondents are in favour. I'd guess that the 31% against were from the other camp: it's a
free poll.
Nemo is gradually getting to be less of a nuisance,
and Yvonne has been training him to do all sorts of things.
Today it was the keyboard:
It's not immediately apparent, but he was doing it on request.
More work in the garden, and got round to further reducing the number of plants on the
verandah waiting to be planted. Planted the area to the west of the eastern garden path,
where the pig face use to be, with
various variegated plants: the smaller ones are apparently a kind
of Agapanthus, but I'm still not sure
what these ones are. The one on the right is the strange plant I picked up
in Napoleons a while ago, which appears
to have some kind of bulb which is now shooting.
Yvonne also planted out some variegated irises. We've had
them for 15 months now, and they've multiplied, but never bloomed. We'll see. Gradually
the area is filling out. Here three weeks ago and now:
Also more plants round the bird bath. I'm hoping that this will end up being bulbs of
various kinds sticking out of the (currently invisible) ground cover:
We had intended to do all sorts of things in the garden today, but the weather didn't agree:
it didn't rain much, but it was relatively continuous. About all I did was finally get a
shopping list together for the things I need to buy to complete the various tasks we have
ahead of us.
A discussion about the merits of Apples on IRC today. It's interesting, not to mention
disturbing, how many people now use Apple, Microsoft and Linux as well as, and in many cases
in preference to FreeBSD. Not that people are
necessarily happier; few have as many problems as I do with “modern” GUIs, no
matter who they come from, but they have a fair amount of criticism too. So what do
we use Apples for? For things that FreeBSD doesn't do well, including hibernation. I use
my Apple mainly for scanning (SANE (which
I claim should stand for “Scanner Access Not Easy”) isn't really a viable
alternative), and for reading SDHC cards
(for some reason my version of FreeBSD doesn't recognize the reader, though that may change
with an update.
Sue Blake suggested that Apple had
good UTF-8 support, so went investigating
that. I suppose it depends on what you want, and in this area I'm really not sure what the
best approach is. There's clearly a trade-off between flexibility and ease of use. In
general most people type in a specific language with a specific keyboard layout; with
practice you can type at up to 8 or 10 characters per second. I can't see a way do that if
you have tens of thousands of characters to choose from.
So what are the options? They appear to be:
-
Switch keyboard layouts. This is the oldest, the easiest to use if you want to write in
only one language and don't need special characters. But beyond providing a good
starting point, it doesn't really address the issues at hand.
-
Compose key: press or hold down a
special key and then enter a number of characters which specify the resultant key. For
example, the sequence Compose T H might generate the single
character Þ. I've had a variant of this for decades as an Emacs macro, and only recently started to use the functionality supplied with
X.
The problem with this approach is that it is relatively limited. How do you input Greek
characters? Maybe it's possible, but then the second problem becomes apparent: almost
total lack of documentation. I've found a number of definition files on my system, with
names like /usr/local/lib/X11/locale/el_GR.UTF-8/Compose
and /usr/local/lib/X11/locale/en_US.UTF-8/Compose. It's not clear which of those
are used, and the man pages are no help. Following what seems to be a trend, there is
no man page Compose(5) on my system, and the web-accessible man
page doesn't say either. It does claim that they're in a very different place,
though.
I've been here before, and once again UNIX timestamps come to my aid:
=== grog@dereel (/dev/ttypk) ~ 85 -> find /usr/local/lib/X11/locale -name Compose | xargs ls -lurt
-r--r--r-- 1 root wheel 0 Mar 26 16:59 /usr/local/lib/X11/locale/C/Compose
-r--r--r-- 1 root wheel 558508 Mar 26 16:59 /usr/local/lib/X11/locale/en_US.UTF-8/Compose
-r--r--r-- 1 root wheel 20411 Jul 28 11:53 /usr/local/lib/X11/locale/iso8859-1/Compose
The references to the UTF-8 variant are old and correspond to my examination of the topic
four months ago. At the time I didn't realize that I
was using the iso8859-1 variant. But that explains the issues I have seen: it
won't allow me to compose UTF-8. But why am I using that one? Compose(5) gives the answer:
The compose file is searched for in the following order:
-
If the environment variable $XCOMPOSEFILE is set, its value is used as the name of the
Compose file.
-
If the user's home directory has a file named .XCompose, it is used as the Compose file.
-
The system provided compose file is used by mapping the locale to a compose file from
the list in /usr/share/X11/locale/compose.dir.
OK, that's doable. But looking at /usr/local/lib/X11/locale/en_US.UTF-8/Compose, I
don't gain much:
<Multi_key> <U10000655> <Arabic_alef> : "إ" U0625 # ARABIC LETTER ALEF WITH HAMZA BELOW
<Multi_key> <U10000654> <Arabic_yeh> : "ئ" U0626 # ARABIC LETTER YEH WITH HAMZA ABOVE
<Multi_key> <U10000654> <U100006D5> : "ۀ" U06C0 # ARABIC LETTER HEH WITH YEH ABOVE
<Multi_key> <U10000654> <U100006C1> : "ۂ" U06C2 # ARABIC LETTER HEH GOAL WITH HAMZA ABOVE
<Multi_key> <U1000093C> <U10000928> : "ऩ" U0929 # DEVANAGARI LETTER NNNA
<Multi_key> <U1000093C> <U10000933> : "ऴ" U0934 # DEVANAGARI LETTER LLLA
<Multi_key> <U1000093C> <U10000915> : "क़" U0958 # DEVANAGARI LETTER QA
<U10000313> <Greek_alpha> : "ἀ" U1F00 # GREEK SMALL LETTER ALPHA WITH PSILI
<Multi_key> <parenright> <Greek_alpha> : "ἀ" U1F00 # GREEK SMALL LETTER ALPHA WITH PSILI
If I read the man page correctly, the <U1000093C> and <U10000915>
are individual keys themselves, so I can't even use these combinations. Even if
I could, it starts getting incredibly complicated.
-
An alternative would be to enter the UTF-8 keycode directly. This seems reasonable, and
I'm sure I've seen it, but I haven't been able to find any way of doing it. Somebody
tells me that Apple can do this if you hold down the option key and enter the
code, but that doesn't work correctly for me: it does create some character, but there's
one for every keystroke, and of course I can't find anything in the Apple documentation.
But it's an area to investigate.
-
Of course, who needs a keyboard when you have a mouse? You can have a keyboard or a
set of characters pop up and just select them with the mouse. That would be
impossibly painful for normal text, but for entering special characters once in a
while, it's bearable. Wikipedia offers this approach when writing pages, unfortunately hiding it where you can't see
it and the text you're editing at the same time. Apple has this too, with one of
its forgettable names: “Character palette”. Went looking for that in
“Mac Help” and got all sorts of irrelevant documents. With a little
help from my friends and a bit of playing around, found that I can configure to use
it, along with a lot of specific keyboard layouts, and select them by clicking
on—wait for it—the national flag at top right next to the other
unrecognizable icons. At least it's possible to use it.
One of the interesting things about the most recent
power failure was that it was clearly more local than most. Even Chris Yeardley had
only a momentary blip, while we were without power for 3½ hours.
Today I found out why: a linesman from Powercor came to ask if they could take the power down for 15 minutes. I said no, if
they want to do preventive maintenance they should inform us in advance. But then I thought
about it and went to talk to him and his mate. On the way I saw something surprising:
That transformer is clearly new. It proved that the old one had failed last Tuesday, and
the 3½ hours were the time it took them to replace it. That all sounds quite reasonable.
It seems that transformers require “tuning” to the correct
voltage—something that must be very difficult given the fluctuations we have
here—and they were dropping it a couple of taps. It proved that they thought they
could do it in 2 minutes, so I gave them the go-ahead, and indeed the power was really only
down for 2 minutes. Things aren't always enough to complain about.
Into town to buy various materials that I need to complete the garden projects: nails,
screws, and some glass clips for the greenhouse:
These clips are made out of thin sheet metal, and it's fairly clear that I couldn't get the
exact component. But it should be relatively simple to cut a strip of metal to size. Went
looking for that all over town—nobody seemed even to have an idea where I could find
something to suit—and finally ended up at Skipton Street Sheet Metal, a rather dubious
looking place without a proper entrance. But the bloke I spoke to had exactly what I
wanted, and he cut me four strips each a little over a metre long in a matter of
minutes—$5. I have the feeling I should have asked for 5 strips.
Part of my driving around took me past the Botanical Gardens, so in to take a look
at what they have at this time of year. Not much by comparison with other seasons, but some
of the succulents had very pretty flowers:
There was also another kind of Aeonium
with many red flowers,
but I couldn't find a label, so I didn't take a photo. Maybe I should reconsider my dislike
of Aeoniums.
Back home, we still have many more flowers. I'm planning to take photos at the end of each
month, so today was the day to do the July page. Some of them are looking less than happy:
But there are still a number that look quite good, and the first of the spring flowering
bushes are following the bulbs:
More investigation of UTF-8 entry on Apple today. Result: it works, once you get past the
lack of documentation. First you need to enable it. You do that by the non-obvious (and
initially invisible) method of setting an Input Method in the System
Preferences Language/Input Menu menu, where you can also enable the Character Palette
(which, I'm told, has changed its name to something else in more recent versions of Mac OS).
The Input Methods look like keyboard layouts:
You select one of them from one of the minuscule, variable icons in the top right of the
screen. But somehow there seems to be a layering violation. On my machine I can select
between Australian, German, Devanagari or Unixcode Hex input:
Unicode hex input works: you hold down the key marked on my machine as alt/option and
type in 4 hex digits. But which input method governs the rest of the keyboard? I don't
know how to tell. Empirically it doesn't seem to be the last selected “real”
keyboard layout: if I select the German keyboard, where the letters y and z
are reversed, and then select Unicode input, it reverts to the non-German layout. What a
mess this stuff is!
Yvonne came back from shopping today with a 1 TB disk and a
USB DVB-T tuner from ALDI. I don't know why I
wanted the tuner: it cost $50, and I've already
established that the Winfast DTV
Dongle Gold ($34) works. Tried this one out. It didn't, and I can't be bothered to
investigate further. Isn't it nice that ALDI takes things back with no questions asked?
So the elections are coming closer, and the parties are promising all sorts of things to
their target audience. I don't seem to be in it, and neither are many people
in Dereel. I've discovered that Dereel is
one of the 52 localities which
have the highest risk of being destroyed by bushfires. On that list it's one of the few
that doesn't have a place of last
resort for people whose houses are burning down. What are the parties doing about it?
Nothing, it would seem, like they're doing nothing about improving communications or
ensuring reliable electricity supply. Today the government released details of the new
National Broadband Network. They'll be installing
initially in areas with good ADSL coverage. Dereel, of course, misses out:
We're in the “satellite footprint” area south of Ballarat. Even that's better
than what the opposition is planning, though: they would scrap the scheme altogether. Wrote
a couple of letters to Darren Cheeseman and Sarah Henderson, and published them
on the dereel.com.au web site. I don't suppose it will do much good. I asked for
confirmation of receipt, and I didn't even get that.
My subscription to Popular Photography takes
forever to arrive. I gather it appears in the USA at the beginning of the previous month,
and when I first subscribed the issues arrived at the end of the month. The times are
getting longer, and the July issue arrived on 18 July 2010.
Subscription renewals are a different matter: today I received a letter by airmail, sent on
28 July 2010, only two days ago, reminding me that my subscription
will soon expire (in December!). And I need to reply by “07/18/10”. I never
cease to be amazed by the deliberate ambiguity people use for dates. After
the Y2K panic you'd think that people would
at least write the year out in full. It can't be to save space, otherwise they wouldn't use
leading zeroes. The 18 in the middle indicates that it's a US out-of-order date—in
this case—but which is the year and which is the day? On the face of it, this date
should mean 2007-10-18, given the leading zero in
the 07, but I suppose it's really 18 July 2010, ten days
before it was sent, and a Sunday. I'm amazed.
We still have rose buds:
I wonder if it will bloom. The last one did, but it looks anything but happy:
Mail from Daryl Tester today discussing slow deletes under Linux. He suspects it has
something to do with file fragmentation, and it looks as if he is right. He points me at a
program called filefrag, which shows the number of fragments in a file. Tried
running it against a number of files that I wanted to delete and compared the deletion times:
| fragments |
|
size |
|
real |
|
user |
|
sys |
|
|
|
|
(s) |
|
(s) |
|
(s) |
| 5 |
|
1374941948 |
|
0.032 |
|
0.000 |
|
0.004 |
| 7 |
|
1478355296 |
|
0.031 |
|
0.000 |
|
0.000 |
| 1688 |
|
2770630260 |
|
3.311 |
|
0.000 |
|
0.036 |
| 3023 |
|
1993571176 |
|
2.868 |
|
0.000 |
|
0.056 |
| 1281 |
|
2719592208 |
|
1.300 |
|
0.000 |
|
0.012 |
| 949 |
|
3203344032 |
|
1.242 |
|
0.000 |
|
0.008 |
| 1470 |
|
2853799768 |
|
3.112 |
|
0.000 |
|
0.028 |
| 95 |
|
6991753652 |
|
0.117 |
|
0.000 |
|
0.004 |
| 1607 |
|
3900287856 |
|
2.857 |
|
0.000 |
|
0.012 |
| 3711 |
|
4198226308 |
|
9.853 |
|
0.000 |
|
0.136 |
| 1560 |
|
2349734920 |
|
3.431 |
|
0.000 |
|
0.028 |
| 2220 |
|
1471019348 |
|
4.807 |
|
0.004 |
|
0.020 |
| 1888 |
|
2851456536 |
|
5.033 |
|
0.000 |
|
0.064 |
| 2437 |
|
3559201148 |
|
4.454 |
|
0.000 |
|
0.028 |
| 22 |
|
4886416288 |
|
0.069 |
|
0.000 |
|
0.000 |
| 3650 |
|
4118056528 |
|
7.592 |
|
0.000 |
|
0.096 |
| 13 |
|
10835008324 |
|
0.024 |
|
0.000 |
|
0.000 |
| 3046 |
|
3636609772 |
|
7.508 |
|
0.000 |
|
0.072 |
| 52 |
|
5344228060 |
|
0.009 |
|
0.000 |
|
0.000 |
| 4888 |
|
6629422004 |
|
10.181 |
|
0.000 |
|
0.100 |
| 2794 |
|
6545229964 |
|
4.472 |
|
0.000 |
|
0.036 |
| 6061 |
|
11617944664 |
|
7.467 |
|
0.000 |
|
0.060 |
| 14 |
|
6181893268 |
|
0.014 |
|
0.000 |
|
0.000 |
|
These files are MPEG-2 transport streams written by MythTV. Why are there such wildly different numbers of fragments? Why, indeed,
should there ever be that many fragments? Is this an indication of poor storage
allocation? I thought that possibly filefrag (written by Ted T'so) might
be ext3 centric, and that it might not
apply to XFS, but these values
strongly suggest that it does.
One of the things I bought yesterday was an L bracket, which I adapted to use for mounting
the camera vertically for panoramas:
I don't need it for all my panoramas, but I expected it to be of use for the verandah
panorama. It required minimal work. Other people who use this kind of approach drill holes
in the correct place to mount the lens over the pivot point of the tripod head, but I don't
need that, since it mounts on a two-way focus rail. Initial tests are encouraging. Here
last week's shot and a test one done today:
Nemo is gradually calming down and behaving, though
it'll be a while before we're done. He's almost making friends with Lilac, helped by the
occasional claw in his nose, but Piccola is a
different matter. Nemo doesn't want to hurt her, but try telling that to Piccola when he
chases her. Today we experienced something unusual: although Lilac doesn't have much
trouble with Nemo, she started growling and chased after him when he didn't leave Piccola
alone. It's not at all cat-like to protect other cats.
I still haven't got round to addressing the subscription renewal reminder I got yesterday.
The “special offer” expired 10 days before the letter was sent, but I didn't
expect the date to be genuine. More to the point was that I don't know whether I should
renew or not. I subscribed 2 years ago for 2 years for the amazingly cheap price of USD
24. This renewal is for 12 months (no 24 month option) and costs $30.
In itself, that's not too bad for 12 issues, but it's 250% what it cost the last time. How
much are the savings? Went looking on their web site and after obligatory navigation (they
don't have a real URL for subscriptions outside the USA) found:
So the “Insider savings” make up a total of -$8. Do they think their
subscribers are stupid? About the only issue I have with taking up a new subscription
instead of renewing is to ensure that the new one starts with the issue following the last
of the old subscription. But even if I overlap up to three months, I save this way. It's
also interesting that I can get the online version for only $2 less than the paper edition,
which includes printing costs and international postage. It's $8 more than the US
subscription, and it requires special reading software: clearly existing software is not
good enough. I'm amazed at the behaviour of these companies.
Yvonne reported a read-only mailbox again this morning. I've
been through that before, and I was able to get
things going again with:
=== root@lagoon (/dev/pts/3) ~ 9 -> /etc/rc.d/statd restart
statd not running?
Starting statd.
=== root@lagoon (/dev/pts/3) ~ 10 -> /etc/rc.d/lockd restart
lockd not running?
Starting lockd.
But why does it happen? It's got to be a race condition, and that's particularly difficult
to catch.
Yesterday's vertical panorama test was done without flash or HDR, and it showed. Today's
was done as I wanted it, and once again I'm quite happy:
I still have a couple of issues:
-
I don't have any way to make multi-layer panoramas. To do that I need to tilt the
camera vertically about the nodal point, and my equipment can't do that (yet).
-
The ball head is probably too weak for the whole setup. It's not designed for such a
lot of stuff to be put on top of it. In particular, the pan head has a rather stiff
vertical adjustment, and it's easy to move the ball head instead when the pan head is
loosened and the ball head is tight.
