Another brief power failure at 7:34 this morning. That's an unusual time, and for some reason it tripped the
non-UPS circuit breaker. Why? There's almost nothing on that circuit.
Peter Jeremy, a Google employee, noted
something that I hadn't about yesterday's problems with Google Maps: I had selected public transport, and that was
clear from the error message. Clearly there is no public transport
between Lauchröden
and Eisenach.
Is it? Isn't it? The error message was:
Sorry, we could not calculate transit directions from "Eisenach, Germany" to "Lauchröden,
99834 Gerstungen, Germany".
What is there in that message that says “public transport”? The only possibility would be
“transit”, but that doesn't have any such meaning, outside Google at any rate. The
OED states:
transit The action or fact of passing across or through a place; a passage or
journey from one place or point to another.
And that applies equally to all the transport methods.
And is there really no public transport between Eisenach and Lauchröden? It's Germany,
after all. And it didn't take long to find this page, which
gives times and prices.
But in one point Peter is right: the bug is related to the choice of transport. Presumably
Google Maps still doesn't have a good understanding of German public transport. And in
passing I discover that I don't have a good understanding of Google's icons:
What does that symbol really mean? Is it supposed to be a train? A bus? I can't tell, but
most likely it's a train. What else does it include? Buses, clearly, but what about boats?
And to the left there's a symbolized car and a right turn symbol. Both seem to give the
same result. I've been using Google Maps for nearly 15
years, and I still don't understand it.
Today I got an email from Effective
Electrical confirming the installation date of the “solar system”: 14/04/19 and
18/04/19. Converting to canonical form, that's Sunday, 14 April 2019 and Thursday, 19 April
2019. Why that?
Ah, sorry, finger trouble. Should read 18 and 19 April. Only two weeks! I can hardly
wait.
In the past we've played
various April Fool's jokes on
each other, but this year, somehow, it didn't happen. We also didn't see any on the web.
There seems to be a general feeling that it's an old, worn-out kind of joke.
Into town today to meet Tomas Kucera and Fyodor Torgovnikov to talk about details of the
solar energy installation—I thought. When I got there, the place was deserted except
for Fyodor Fred, who showed me a display of the two kinds of inverters that Tomas
sells:
The choice is either the two on the left or the one on the right. As I had guessed, Fyodor
represents Ingeteam (the one on the right),
but since I had already decided for them, that wasn't a disadvantage.
Much discussion, to the point that I was nearly late for my next appointment. Managed to
get a feeling for how the inverter communicates with the world, via
the Internet Protocol but not
necessarily via the Internet, as they
put it, using 802.11. It seems (I didn't
follow through) that one mode of operation is to go via the Ingeteam web site in Spain. I
wonder what kind of confusion gave rise to that model. Maybe to make it possible to access
the data from a remote mobile phone? And there was certainly confusion about the IP
addresses. He was able to communicate with the thing using his Microsoft “Windows” 10 laptop, a horrible thing that doesn't
have a scroll function for the mouse substitute; instead you have to use the touch screen.
It seems that the inverter uses the same RFC3927 range address that the Selectronic uses, in this case
169.254.1.1. But somewhere I saw an option to
use DHCP, so hopefully I'll be able to sort
that out.
And the second battery set? That's from the battery supplier, not Ingeteam.
They're hoping for a supply in October, but Fyodor didn't sound overly optimistic.
If for some reason they can't supply it, they can always take the existing battery back and
install something from another supplier instead. It's interesting to note that the inverter
uses high voltage Lithium-ion
batteries, 256 V, but in fact it will run on anything with the correct voltage, including 21
lead-acid batteries. And that's how the defaults are set up, allowing for only about 50%
discharge:
One thing that he couldn't answer—surprisingly—was how to configure the thing to not
discharge the batteries below a certain point as long as grid power was present, but to
allow use of that charge when the grid fails. He noted that others had asked that too, and
in principle the setup suggests that it should be possible. I got a mail message from him
later in the day to confirm that yes, indeed, it's possible.
Left Fyodor in a bit of a hurry to get to my physiotherapy appointment, and in fact made it
just about on time. Things are looking better, and Heather gave me some exercises to do
that should be more useful than the ones I had in the Dereel Hall, including one of dubious
promise where I need to sit on a roll and move around:
The characters are set strangely, but I read Cari laksa. Cari laksa? The Malay term,
written elsewhere, is “Kari Laksa”. In Malay, the letter C is now pronounced like
a ch: chari laksa, which coincidentally means “Look for (the) laksa” in Malay. I
wonder why they did that.
And the results? It's OK. It's surprising how little difference there is between the
individual brands. Next time I'll try another one that looks decidedly dubious: apart from
only having half the quantity, it wants condensed milk in the sauce. I don't think
I'll even bother trying that ingredient.
For reasons I don't understand, my Android phone
(taskumatti.lemis.com) keeps waking up, turning on the display, sometimes making a
noise. It's not annoying enough that I have bothered to try to find out why. It does
enough more annoying things.
Today, though, it excelled itself. Noted a display “NOKIA” on the screen (it is, after all,
a Nokia 3). And it alternated with
various error messages, tastefully not left there long enough for me to get really worried.
One suggested that Android had crashed, but that was clearly so worrying that it didn't come
again. The others looked like this:
How about that, a real crash. There must be some way of resetting the thing. In the Good
Old Days, removing the battery would have done it, but this has a non-removable battery.
OK, off to the web to look for wisdom: reset hung android phone.
Ah, “hung” is an old, worn-out (if strong) magic word. Nowadays it's “hanged”, clearly a
substitution made by some modern person who doesn't understand the difference between
transitive and intransitive verbs, nor the difference between weak and strong verbs. But I
found a solution anyway: Vol- and PWR together.
Problem: it didn't work. Phone seriously braindead? Tried the search again with the
mention of the phone model, and came up with this page: press Vol+ and PWR together. And that worked, though not
quite as described.
Huh? What stupidity could lead to such a flagrant violation
of POLA? You'd almost think they were
trying to annoy people. And yes, the suspicion of brain damage wasn't that far from the
mark, though it didn't relate to the phone itself. But later I realized that there was a
different issue at hand. Vol- and PWR now has a different meaning: take a
screen shot. I wonder when they'll change it again to mean “perform factory reset, no
questions asked”. So yes, braindeath and POLA violation, just on whose part?
The phone came back up, but clearly some things weren't the way they should be. In
particular, Yvonne called from town, and not only did I not
get my custom ring tone, I got no ring tone at all, and the phone didn't recognize the
number. Further investigation showed that it had forgotten my entire contact list:
OK, sign in. This horrible glass keyboard! And I had to go looking for my password,
which was suitably complicated. The usual messing around, then it asked me whether I wanted
to use the account I had signed in with, or a different account. Why? Pressed “Next” to
select the account, and got:
I've seen that before. And I saw it again, and again, and again. I couldn't find any way
to sign in, nor any reason why I was taken back to the login page (of course Android is too
polite to give reasons). I did, however, get email from Google:
What does it mean, a new Nokia 3? It probably knows more about the device than I do.
About the most interesting thing is that it now knows roughly where I am; previously it had
me located up to hundreds of kilometres away.
At least the explanation is different. OK, it's my guess that I'm not the first person who
has been bitten by this. What advice do I get on the net? Many and varied, including the
claim that I have disabled my Google account (wrong; it works everywhere else) and that I
should delete the account and recreate it. And then suggestions that I should clear cache,
along with incorrect paths through the maze of twisty little menus (presumably correct for a
different but unstated version of Android).
Finally found out how to clear cache on my (Android 8.1) phone, rebooted (this is
Microsoft, after all, isn't it?). No change.
OK, 30 minutes down, no closer to resolution. The cache issues, even if they're not what's
biting me, point to bugs in Google's applications, and there's no obvious will to fix them.
It would be quicker to reset the device to default and set up again. Did that.
So I reset the phone and rebooted it. This time it seemed a good idea to write down the
steps.
First, even before anything else, it asked me if I wanted to restore from the cloud. OK, if
that works, it's probably the easiest way. But first it needed to connect to
the Wi-Fi network, which, for some reason,
took several minutes. Why? And then it found:
OK, that looks good (and makes me wonder why it thinks that this device is an “unknown Nokia
3”). Selected that and got quite a reasonable selection of things to restore:
Why Telstra stuff? It can't have been bundled with the phone, which was supplied locked
to Vodafone, and I certainly didn't
install it. So why “Included”? And how so I say, “No, thanks, not with a bargepole”?.
Somewhere round here it wanted to know a phone number and get me to agree to the service
terms—why now? And off it went with the information that I could continue setup.
Next was to set up the “voice assistant”. Oh, no, sorry, you've already done that. And in
fact there wasn't much more to do except set the phone ring tone, not for the first time.
That was more complicated than I thought. It seems that I had downloaded them to the
directory /storage/emulated/0/Music/, which is what's well hidden somewhere on the
menus (I had expected something as simple as “Music”, but I don't have that; maybe I need to
spend money to access it from the GUI). But in this case the directory was empty. It seems
that I had had a problem with the microSD
card:
Why that? It was working before. And why is it called “Portable storage”? What on
this device isn't portable? OK, set it up again, and in contrast to last time (where it
hung), this time it worked. OK, now upload the tones again.
And at this point the X display on eureka
hung while trying to display the directory on the phone. Switched to the other server and
shot down one firefox after
another, before finding the real culprit:
USER PID %CPU %MEM VSZ RSS TT STAT STARTED TIME COMMAND
grog 69481 0.0 0.1 125240 33928 v0 S 3:00pm 0:00.01 /usr/local/lib/nspluginwrapper/i386/linux/npviewer.bin
Finally I was able to download the tones and confirm that the settings menu automatically
finds them in that location. But what else uses them? All I can find are things like “Play
Mus..”, which wants me to pay money to look at something:
Sooner or later we'll finally have all the electrical updates done. One question that's
open is whether we need to mount another flash unit somewhere to illuminate the kitchen
better. One possibility would be on the corner from the hallway to the lounge. Today I
tried that with a studio flash to match the existing unit on top of the (from this vantage
point) left-hand fridge:
How do I match the illumination? Both units are adjustable, but how? The existing one, a
Godox Smart
300 SDI, with a maximum output of 300 J, has 6 settings, marked simply with bars (or,
as the web page puts it, “LED Precise Power Display”, “8 steps”). And the test one, a
Godox DE 400 (already a couple of years old and
thus disowned by its maker), has a maximum output of 400 J and a scale reading from 5.0 to
7.0 in steps of 0.1. It seems that 1.0 on this scale represents
2 EV (why do they do these
things?), so it seems that the 0.1 steps are steps of the tenth root of 4 (1.148, or 0.2
EV). So how do I set 300 J output? 0.4 EV, it would seem; that gives about 303 J.
Tried that and measured illumination in various places with five combinations. It's not
quite clear what the bars mean on the old flash, and it's difficult to get precise readings
because of the shadows I cast (and which I need to take into account when taking photos),
but here are the results:
New flash 300 J. Aperture setting range 2 EV (f/6.9 to 13.5).
Old flash 5 bars. Aperture setting range 1.4 EV (f/4.44 to f/6.8).
Old flash 5 bars, new flash 300 J. Aperture setting range 1 EV (f/10.2 to f/14).
Old flash 6 bars. I didn't measure all positions, but in some places it was f/4, not
enough illumination, and in other places f/10.6, a range of 2.6 EV.
Old flash 6 bars, new flash 200 J. Aperture setting range f/10.2 to f/12.2 (0.6 EV).
Clearly version 5 was the best, and it gave relatively even lighting:
Only the critical area round the stove was slightly underexposed, but that's really only
visible before postprocessing. Here exposures at f/11 and f/8 before and after
postprocessing:
So I think I can mount a flash unit there. It looks as if 200 J should be enough, and with
that I can use f/9 or f/10 at 200/24° ISO, so the first stove photos was ⅓ EV to ⅔ EV
underexposed anyway.
Yvonne bought some chicken skewers for experimenting with our
newest “Air fryer”. It comes with a
“skewer rack”, a rod with two plates for mounting each end the skewers.
It also comes with metal skewers for use with the rack, but our skewers were already on
wooden, well, skewers. Would they fit? Yes, with unrelated issues:
The plates hold 10 skewers, so they're at 36° intervals. The central shaft has a square
cross-section, and it's possible to mount the plates in any orientation. So I mounted them
90° apart, giving an 18° offset between the holes at each end. “Well Don't Do That Then”.
Then there was the issue of getting the thing into the “fryer”:
They were cooked, but only just, and not in the slightest browned. But the results tasted
very good, and possibly this is a reason to keep the thing.
But why so pale? This was chicken, and the pieces were smaller than recommended. How would
it work out with larger pieces of beef? I'm sure that things would be much better at a
higher temperature. That's easy to do, of course, and we will next time, but once again I'm
left wondering who comes up with these default settings.
To my surprise, I discovered that we've now been in Stones Road for half the time that we
were in Kleins Road (here from 7 May 2015, in Kleins Road from
10 July 2007 until, well, 7 May 2015). You wouldn't have noticed
from the garden, which still looks pretty terrible. But the flower boxes in front of my
office are a mixed bunch: on the one hand I planted a ground cover there in the spring and
watched it die, but it still has lots of bulbs in it. Soil problems? Today Mick the
gardener came, and I got him to replant the one I suspect of having particularly poor soil.
There were certainly enough bulbs in there:
Also removed a number of Mirabilis
jalapa plants, some of which ended up in the north garden. At least they grow well with
no attention. Petra Gietz took some of the bulbs with her, and so did Mick, and he planted
more in the rose garden, leaving the rest behind.
Heather has given me some gymnastics to strengthen
my gluteal muscles. One of them
involves standing with one foot in front of the other and with eyes closed. That seems
straightforward enough, but it's surprisingly difficult to keep your balance. I've taken to
doing it when cleaning my teeth, but today something went wrong, and I found myself in the
bath. Did I lose consciousness? No idea. I don't think so, but there's a fraction of a
second when I'm not quite sure what happened. She was clearly right when she said that I
should take precautions. Still, apart from a scratch on my shoulder and a bit of minor
bruising, everything seems to be OK.
It's been 11 days since I
planted my Epazote seeds. When
are they going to germinate? Even parsley only takes a week. I've been looking carefully
for some days, then this morning I saw some seedlings which look dried out.
Why? I've been watering them constantly. Dragged out my Olympus STF-8 flash unit and a macro lens and took some photos:
That doesn't look that bad after all. They're just taking a very long time to germinate;
they're still less than 1 cm high. By contrast,
the Kniphofia seeds that I planted some
time later have all germinated and are now about 3 cm in size:
I've had my Olympus STF-8 flash unit for over a year
now, and I've hardly used it: it's such a pain to put on the camera. When I did so
today to take the photos of the seedlings, the batteries were discharged.
Discharged? Pretty much dead. All of them showed a voltage of 0.3 V or so. They're
all Nickel-Zinc (NiZn)
batteries, which I've been experimenting for over 8 years, and they should have a voltage between 1.6 V (discharged) and 1.8 V
(charged). What happened? Put them in the charger. One just died, but the others charged
again.
My experience with NiZn has been that they discharge unevenly, so the voltages are very
different on discharge. But the even discharge of these, especially after so relatively
short a time, suggests to me that the STF-8 might be trickle draining them. I'll have to
leave them out when I'm not using it.
Can you grill Bratwurst in
an “air fryer”? Of course not. It's
not a grill. But it has illusions: there's a “grill” setting on the control panel, which
translates to 12 minutes at 200°.
That's a big difference from real grilling, where the rule of thumb is one minute per mm
thickness. In practice I find that mine (28 mm thick) take about 20 minutes. Still, worth
trying. Here before and after:
Once again there's not much difference to be seen. As with the skewers on Thursday, they're very pale, and
they were just cooked enough. But again as with the skewers, they show promise.
In passing, it's interesting to note how the sausages “sweat” fat:
While walking the dogs down Harrisons Road, saw a number of people and bikes in the
distance. They appeared to have dogs with them, and Yvonne got the impression that they were about to start a race. OK, it would make a good photo,
and if they misbehave, we have evidence:
No, no misbehaviour there. Yvonne thought that it might have been a family outing. But
what got me the most was that the bikes made almost no noise. Why are the typical bikes
round here so noisy?
In fact, the commit came before the error message. What went wrong there? I had no reason
to change that line, but it seems that I had done. Why? Reversion to an older version? I
have some vague recollection of issues with capitalization of port names, but I can't see
anything in my diary. And why didn't I notice?
And then, why is graphics/OpenEXR wrong? It seems that the name of the port has
changed some time in the past couple of years:
=== grog@eureka (/dev/pts/16) /usr/ports/graphics 16 -> l -d /home/src/FreeBSD/svn/ports*/graphics/OpenEXR drwxr-xr-x 4 grog lemis 512 23 May 2017 /home/src/FreeBSD/svn/ports-rot0/graphics/OpenEXR
=== grog@eureka (/dev/pts/16) /usr/ports/graphics 17 -> l -d /home/src/FreeBSD/svn/ports*/graphics/openexr drwxr-xr-x 3 grog lemis 512 28 Dec 04:03 /home/src/FreeBSD/svn/ports/graphics/openexr
The normal problem, of course: the fryer is too small, and we were left with a couple of
pieces that didn't fit. Barely cooked them in the microwave oven, then grilled them for a
couple of minutes. It's difficult to tell the difference:
With tanduri chicken goes tanduri nan, of course. We cheat and buy partially cooked “naan
traditional” from ALDI and heat it in the roti
cooker. Problem: the cooker isn't big enough. Today I discovered an old sandwich toaster
that Yvonne used to use. Looks like just the right thing.
And how about that, it did the job.
Yesterday's success with the
tanduri nan made me think: I've been cooking my tortillas in the roti maker, and I've had
difficulties: either they dry out or they don't get cooked. Can the sandwich toaster do it
better? At any rate it has the advantage that the top has a double mount, so it doesn't
produce wedge-shaped roti or tortillas the way the roti maker does. Tried that today.
First question: how close do the two surfaces get? A distance of several millimetres is OK
for a sandwich, but too much for a tortilla. Tried it with a chopstick in between and
decided that it would probably be OK.
Not only were the two surfaces close together, the weight of the top squeezed the tortilla
and made it wider, in the process tearing it apart. I'll have to think of how to avoid
that; probably wait until it's partially cooked before closing it.
The results were OK, though one part was a little tough. But it's something to experiment
with.
After last week's experiments with flash in the kitchen, it's time to buy a new flash head. That's
straightforward enough; go to eBay and find
what they have. And I found one that looked surprisingly cheap:
Only a little over $100 for a Godox DS400II, along with a flash trigger! Looking at the
photos is even more exciting:
TTL flash control! And it's available for Olympus! Just what I always wanted!
Hah hah, only joking. For $104 you get only the flash trigger. If you want the item
advertised, you first need to select it (another $237). And if you want both, you have to
pay both prices, a total of $341. And if you read very carefully further, you'll
discover that you also need a receiver, which they don't even offer.
Sorry, eBay seller qlians_au, you're not trustworthy. I'll read your descriptions,
but I'll never buy anything from you. It's a pity that eBay hasn't stopped this kind of
deception.
Still, does this really work? More investigation shows that Godox' web site is appallingly inconsistent, and though I
found the documentation for the XProO (PDF), I
couldn't find an up-to-date list of which models it supports, nor which of them work with
TTL flash. My best guess so far is: some models have built-in receivers, others have
triggering via an external receiver and a USB port, like the DS400 II for sale here, and
others again, like my DE400s, have only a slave receiver and a conventional sync cable.
Clearly you can't use TTL without a receiver. And clearly the built-in receivers support
TTL. But what happens with USB? My guess would be that USB wouldn't be fast enough to work
with TTL. On the other hand, if not, what is the advantage? I have really cheap
remote triggering hardware (transmitter and receiver about $15 together). Why use expensive
stuff and USB connections unless it can do more?
As if that wasn't enough, it occurred to me how similar the DS400 looks to my DE400:
So: how do I find out if it supports TTL mode? Is there a (shudder) Facebook group? Yes. And of course I have to be
approved. Applied, but didn't get any results before evening.
Last week I tried the
“Dollee” laksa paste and found it barely distinguishable from the “Teans” laksa paste that I
normally use.
Until today, when I ate the final portion. It was filthy! Well, no, but that was the first
impression. Some of the spices (pepper?) clearly weren't ground enough, and they left a lot
of sediment in the bottom of the pot. Nothing serious, but given that Dollee doesn't seem
to have any advantages, enough reason to stick with Teans.
Since moving to Stones Road, we don't have a convenient circular route to walk the dogs: we
have to go up to a point and then turn round and return the way we came, usually along a
road. We could walk up to Progress Road to Rozenstein Road and then back via Bliss Road,
but that's 3.3 km, rather more than we really want. We did have a way that cut across
through bush from Progress Road to Bliss Road about halfway up, but that seemed to annoy the
people who lived on the neighbouring block, so they blocked the route.
In fact, one possibility would be to go 300 m down Bliss Road and then along the back of the
blocks facing Stones Road to the back of our block. Took a look today. No:
In the past I've noted that my fisheye
lenses have particularly little flare. Today a question came up on Quora: “Do certain
cameras have better lens flare, if so which ones?”. Well, I wouldn't call any flare
“better”, but certainly it was worth comparing a fisheye (the Olympus M.Zuiko
Digital ED 8mm f/1.8 Fisheye PRO) with my widest rectilinear lens, the
Zuiko Digital ED 9-18 mm f/4.0-5.6, with shots directly into the sun. The results
weren't as pronounced as I had expected:
A while back I set journeled soft updates on teevee:/, which caused a rather
unexpected problem:
# dump -2uf - / | nice bzip2 > /dump/teevee-FreeBSD/2/root.bz2
mksnap_ffs: Cannot create snapshot //.snap/dump_snapshot: /: Snapshots are not yet supported when running with journaled soft updates: Operation not supported
dump: Cannot create //.snap/dump_snapshot: No such file or directory
What do I do there? I can hear “Use ZFS!”
ringing in my head, but the obvious thing is to disable them again. But how? Tried today
and ran into unexpected problems:
So I've gradually come to terms with the wine that I bought online last month, and in the meantime the
seller (McLaren Vale Cellars) has
been sending me direct mail offering me what look like good choices of wine. OK, let's try
some Shiraz. Put a dozen bottles in my
“Shopping Cart” and try to pay for it:
OK, that's simple: do what it asks. But I couldn't! The field is set not to accept pastes:
it wants me to enter the email address manually! Clearly a bug, so I selected the
“Contact Us” page and had to jump through more hoops before I could finally send a message.
And the reply I got really blew my mind:
Unfortunately our system makes you type it in .
Why? Why don't they get it fixed? Don't they want to sell anything? To quote Daniel
O'Connor on IRC:
<Darius> groggyhimself: user interface sadism
For me the decision was borderline anyway, but that has made the decision for me. There are
other companies out there. Goodbye McLaren Vale Cellars.
Petra Gietz along with her granddaughter Kahlan (pronounced Kehlen) this morning to pick up
some of the Carpobrotus that was overflowing the garden bed—about the only thing that has
really grown well in the 2½ years since we planted it. Here when we planted it and now:
In principle I'm relatively satisfied with my Saeco coffee machine that I bought 3½ years ago. As I noted at the time,
the “documentation” is a catastrophe, and I won't forgive Philips for not giving my $50
cashback.
But even with that, it has probably made itself paid. Previously I had had a capsule
machine that cost $0.375 (itself a good price) per cup. With fresh coffee that drops to
about $0.09. For 4 cups per day, that makes a good $1 a day, so the $650 I paid for it were
paid off well over a year ago.
But it has another really irritating “feature”: automatic descaling. What for? We use
rainwater, which contains no minerals. How do I turn the ”feature” off? Ah, can't do that,
might damage your machine. So every few months I'm greeted with this display:
OK, how does that work again? RTFM? That
way madness lies. Follow the menu? Potentially less mad. OK, fill the water container,
put a pot under the output, and point the nozzle into it:
It used no less than four containers of water—more than I recall from previous occasions.
On each case I had to swivel the nozzle to one side, giving it a chance to piss on the
floor. At the end, after it had said “done” and asked for confirmation, it once again
pissed on the floor.
My GPS navigator was behaving strangely today. The trip log
to Ballarat showed the wrong date, 26
August 1999. And the time was off by a couple of hours. Hardware issues? Restarted the
thing without any improvement.
Back home, discovered that it had had its Y2K
event: the GPS week count had passed 1024, and so it had gone back 1024 weeks in time. That
doesn't explain the 2 hour time discrepancy.
But what were the GPS designers smoking to limit themselves to 1024 weeks, somewhat shy of
20 years? Were they expecting the technology to be replaced by then?
Who's the most well-known BSD
personality? Jordan
Hubbard? Theo de Raadt? That
depends on whether he's known for BSD or something else. From
the NetBSD project we have proff,
now better known as Julian
Assange. And he's certainly in the news. What got me is how old he looks:
He almost looks older than I, but he's only 47, young enough to be my son. I wonder what
went on in the Ecuadorian Embassy. And will they reckon his 7 years' self-imposed
imprisonment to whatever sentence he now gets?
While discussing proff's connection with BSD, went looking through the commit logs.
He seems mainly to have been involved with a package
called Surfraw, which he wrote:
Module Name: othersrc
Committed By: proff
Date: Mon Jan 15 07:57:23 UTC 2001
Update of /cvsroot/othersrc/surfraw
In directory netbsd.hut.fi:/tmp/cvs-serv10128
Log Message:
initial import of surfraw-1.3.0
Surfraw provides a fast unix command line interface to a variety of
popular WWW search engines and other artifacts of power. It reclaims
google, altavista, babelfish, dejanews, freshmeat, research index,
slashdot and many others from the false-prophet, pox-infested heathen
lands of html-forms, placing these wonders where they belong, deep in
unix heartland, as god loving extensions to the shell.
For example:
$ ask why is jeeves gay?
$ google -results=100 RMS, GNU, which is sinner, which is sin?
And then there are a number of
old fortune data files entries, finding
things like:
NetBSD - free yourself from all Stallmanist thought!
-- Julian Assange
Clearly a bridge builder. Some of the “fortunes” in netbsd-o seem offensive even to me; I
wonder what a lawyer would make of them if the rape case in Sweden ever goes ahead.
Victorian Government: we don't need no steenking solar power
Call from Tomas Kucera of Effective
Electrical in advance of our solar panel installation next week. Surprise, surprise!
The Victorian Government has cancelled the solar energy installation rebate with immediate
effect! I can now pay $2,250 more out of my own pocket!
Clearly there's no election coming up. But what are they thinking? It seems that they will
reinstate it on 1 July, maybe. Or maybe I can get $4,000 for adding batteries to an
existing installation. I can do that, but will they honour their obligations? Grrr!
This weekend Yvonne has organized a horse clinic, with Anke
Hawke teaching. Yvonne had arranged that Anke will be staying here, and also Terry (surname
unknown). That required tidying up the “library” opposite my office, which I had been using
as a junk store, so that Terry could sleep there.
And then Yvonne got a message: Terry had had an accident, a tree trunk fallen on his back.
We're still not sure exactly how bad it is, but it seems that he will recover. Clearly,
though, he's not going to be doing much riding in the immediate future, so only one visitor.
Mick the gardener along this morning for more bulb transplants and odds and ends. For once
there was barely any wind, ideal weather for spraying. Out to pick up the sprayer
and... the battery was empty. Yes, I should have thought of that a couple of days ago. In
fact, I did think of that, but that's as far as I got.
OK, find the charger, put it on charge. How much charge can I get in 90 minutes? Tried it
out, and how about that, there was enough for a full load of spray. In fact, we ran out of
weed killer first, after spraying 3 tankfuls of spray. I wasn't expecting that.
Later in the morning Petra told me that she had heard a strange noise (click?) from my
office, and thought that it might have been a power failure. Checked, and we were off the
net! I didn't see any lights on
the NTD. But while investigating, it
was back again! No power failure after all, just a brief spell off the net.
Whose fault? As time went on, that became clearer. We had no less than 7 failures, a total
of 14,221 seconds or nearly 4 hours:
Start time End time
Duration
Badness
from
to
(seconds)
1555032141 1555032624
483
0.003
# 12 April 2019 11:22:21
12 April 2019 11:30:24
1555034739 1555035952
1213
1.702
# 12 April 2019 12:05:39
12 April 2019 12:25:52
1555037101 1555037416
315
3.133
# 12 April 2019 12:45:01
12 April 2019 12:50:16
1555038451 1555048766
10315
3.478
# 12 April 2019 13:07:31
12 April 2019 15:59:26
1555049372 1555049396
24
5.941
# 12 April 2019 16:09:32
12 April 2019 16:09:56
1555053989 1555054531
542
0.784
# 12 April 2019 17:26:29
12 April 2019 17:35:31
1555055199 1555056528
1329
5.389
# 12 April 2019 17:46:39
12 April 2019 18:08:48
Date
Outages
Duration
Availability
Date
(seconds)
1554991200
7
14221
83.54%
# 12 April 2019
OK, time to download the Aussie BroadbandAndroid app
and painfully type in the passwords (which were not intended to be typed on glass), checking
each individual letter before it turns into a pumpkin. And then I was able to lodge a
fault. First, though, it had to check my connectivity. Success!
Success? What kind of nonsense is that? The NTD was clearly showing an ODU (outdoor unit)
fault. There was no possibility of establishing connection. What a waste of time!
Entered the fault, and almost immediately I had connectivity again, so (with a real
computer) reported the fact. But then it went down again. Tried calling up and was
informed that I was in for a 21 minute wait. And that with expensive mobile phone calls!
But at least that indicated that I probably wasn't the only one.
Tried to enter the fault again. Hang on “Analysing”. But then I discovered that I had run
out of mobile credit. Simple, buy some more. How? I was doubly off the net. In the end
got Yvonne to buy a voucher when she was in town.
Our Borzois may be big, but they're not
exactly aggressive. In fact, they're far too friendly for any protective activity. But
when Anke Hawke arrived this evening, I heard a loud bark and howl from the (dark) hallway.
Had she brought a dog with her? No, it was Nikolai! I can't recall having heard him bark before. It seems that he found it
highly suspicious that anybody should come into the house in the dark. Once he saw that it
was Anke, he was happy.
I suppose that's a good sign. If anybody comes into the house when we're not there, I don't
think Niko would be as easily placated, and an aggressive-looking dog of his size would
scare most people off.
While in town on Thursday, I bought a couple of saucepans and frying pans at ALDI. They were cheap, and in contrast to the old ones
that we have, they are intended for use with induction cooktops.
This morning I put one on the cooktop and turned on. Not recognized. One of the pans
worked, but heated very slowly, and I had to use another one.
Why? They looked OK. But looking more carefully at the saucepan, the reason became clear:
The dotted pattern on the bottom is one of the kinds of induction-compatible bases that are
in common use. It's only 10 cm across, with a 5 cm gap in the middle. The gap in the
middle is of no importance, since the induction coils don't heat there, but effectively this
has made a 10 cm pan out of a 16 cm pan (rather less than 30% of the surface area), and none
of my induction fields can talk to anything that small.
And the other pan? It's 30 cm in diameter, just what I need for bigger things. But the
heating surface is only 17.5 cm in diameter, again only a third of the area. By contrast
the existing 26 cm pan, also from ALDI, has a different heating surface 23 cm in diameter
(78% of the area):
Taking the photo of the bottom of the saucepan wasn't easy. It took me five attempts. The
first time (once again!) the flash didn't trigger, and the next one, taken with TTL flash
and standard settings, looked far too dark:
While processing my house photos this
afternoon, the screen suddenly went black, to be replaced by a constant stream that I could
only decipher by taking a photo:
Ugh. "vm_fault: pager read error" doesn't look good. Some things still seemed to be
working (the weather application beeped to say that it
had once again crashed), but clearly it was reboot time.
ada0 is the first disk, the one with the root partition /dev/ada0p2. Without
a root file system I'm dead in the water. But why the detach? There were no previous
messages. The error = 6, on the other hand, is completely
understandable: “Device not configured”.
Rebooted and considered whether this wasn't forcing my hand to finally upgrade the
system. Yes, indeed: it happened again only a couple of hours later:
OK, nobody's home (Yvonne and Anke were at Chris' place with
Anke's clinic), so I had several hours to migrate to the new system. I needed them, and
more.
First I discovered that I still hadn't enabled journaling on /home
and /Photos. Setting on /Photos was strange:
=== root@eureka (/dev/pts/3) /home/grog 58 -> tunefs -j enable /Photos Using inode 321 in cg 0 for 33554432 byte journal
That's reading something from disk. What? And why? Finally it completed:
tunefs: Journal file fragmented.
tunefs: soft updates journaling set
Is that serious? Looks like another UTSL
moment. My guess is that tunefs wants to put the journal file somewhere specific
with contiguous block locations, but that it couldn't find anything suitable, and that
noncontiguous locations might have a slight performance impact.
/home also came up with a couple of strangenesses while running fsck:
Those are the same two files. The first run had set EOF values that the second run didn't
like, so it had to be done again.
Booted into the new system, which was painless enough, but X wouldn't start: “no screens found”. I've been fighting X configuration for nearly
30 years now, so it wasn't exactly unexpected. OK, check the /var/log/Xorg.0.log: I
was using version 304.137 of the nvidia driver. Which version did I need?
For that I needed web access. That brought me back to where I was three weeks ago:
install BIND. This time I had made sure to
add the external name server to my /etc/resolv.conf. pkg search is my friend.
Is it? “Permission denied”. Build from source? First I had to find it. What's the port
called? What directory?
OK, without X it was difficult to do much, so went to teevee and tried from there.
Ah, BIND is so important that it's not under net or net-mgmt or such,
but dns. And there I had
Which one? Why isn't there one just called bind (possibly with a symlink to make
things clear)? bind-devel suggests that it's a little too bleeding edge. What
about bind914? Checked bind914/pkg-descr, no warning about being unstable or
whatever. Back to eureka, tried to build. Bloody EPERM again!
Back to teevee, did a make fetch. Success. Back to eureka:
===> bind914-9.14.0 depends on package: py36-ply>=0 - not found
===> License BSD3CLAUSE accepted by the user
===> py36-ply-3.11 depends on file: /usr/local/sbin/pkg - found
=> ply-3.11.tar.gz doesn't seem to exist in /home/src/FreeBSD/svn/ports/distfiles/.
=> Attempting to fetch http://www.dabeaz.com/ply/ply-3.11.tar.gz
fetch: http://www.dabeaz.com/ply/ply-3.11.tar.gz: Permission denied
=> Attempting to fetch http://distcache.FreeBSD.org/ports-distfiles/ply-3.11.tar.gz
fetch: http://distcache.FreeBSD.org/ports-distfiles/ply-3.11.tar.gz: Permission denied
=> Couldn't fetch it - please try to retrieve this
=> port manually into /home/src/FreeBSD/svn/ports/distfiles/ and try again.
Dammit, back to teevee and fetched it. What's BIND doing
with Python anyway?
Finally got BIND built, watching it joyfully print:
* THIS IS A DEVELOPMENT VERSION OF BIND, IT MAY EAT YOUR DATA *
So is it stable? Isn't it? I have more important things to worry about.
But after starting it couldn't resolve anything. The overly voluminous log messages told me
that I didn't have any zones. Grrr. I've been here before: BIND has had a long history of
migrating its configuration files, from /etc/namedb/ to /var/named/etc/namedb/
and (currently) to /usr/local/etc/namedb/, where it wanted them now. OK, put the
info there, remember (forcedly) to change /etc/namedb to /usr/local/etc/namedb
in /usr/local/etc/namedb/named.conf, and finally it was working.
Round about here (rather late) it became clear that the EPERMS were due to firewall
issues. For some reason I had in my ipfw rules:
00040 1055 44252 unreach filter-prohib tcp from not 192.109.197.0/24 to any setup
How did that ever work? Removed it and finally had connectivity.
Gradually time was getting on, and my main concern shifted to giving other computers in the
house access to the outside world. It seems that squid wasn't doing what it should,
so I had to set the browsers on teevee and lagoon to connect directly to the
Internet, and also had to mess around with email, which hadn't come back the way I expected.
And MySQL and httpd weren't doing what I want either. The latter somehow got
started with the flag -DNOHTTP. Are they now mandating https? Where do I
get a valid certificate? Dammit, this is all more pain that I need.
In the evening—Yvonne had outdone herself and taken 24 GB of videos and photos—made a photo
backup. That normally takes about 20 minutes. Today it took 2 hours! Why? Something to
do with my file system parameters? One of many questions to put off until tomorrow.
Dinner with Chris Bahlo (as usual on a Saturday) and Anke Hawke this evening, Anke noting
that she had a tendency to leave her wine glasses half full. I promised to write it up in
my diary, but when I checked I found that she had drunk them after all (or maybe Yvonne, trying to keep the peace, had poured out the rest).
Piccola must have tried them out. The pill
that she bit through (and obviously found not to her taste)
was Allopurinol, which doesn't seem
to be particularly toxic. And she doesn't seem to have swallowed any of it, so hopefully no
harm was done.
Started pretty early in the morning trying to get eureka back up and running. Given
the issues I had, started copying the old root file system to the spare /destdir
partition on the new disk, so that I would have the files to compare, and if everything went
wrong I could boot the old system from it.
The obvious next thing to do was to get X up and
running again. Which driver version did I need? That's described on the list. I have a
GeForce 9500 GT and a GeForce GT 640. Which version of the driver do I need?
That's straightforward enough: this page, with the rather misleading title “What's a legacy driver?”, tells me that
I need driver version 340.xx for the 9500 GT, and driver version 390.xx for the GT 640.
Huh? There can only be one, and it used to work. Is this only half the story, or have they
changed things so that I can't run both cards with the same driver?
Ok, let's try. Deinstall the current driver and...
===== Sun 14 Apr 2019 09:33:49 AEST on eureka.lemis.com: pkg install nvidia-driver-340-340.107_3
Updating FreeBSD repository catalogue...
FreeBSD repository is up to date.
All repositories are up to date.
The following 905 package(s) will be affected (of 0 checked):
Installed packages to be REMOVED:
libabw00-0.0.2
...
ImageMagick-6.9.5.7,1
a2ps-4.13b_8
samba42-4.2.14
emacs24-24.5_3,3
autopano-sift-C-2.5.1_5
OpenEXR-2.2.0_5
php56-5.6.26
php56-session-5.6.26
php56-mysqli-5.6.26
php56-mysql-5.6.26
...
New packages to be INSTALLED:
nvidia-driver-340: 340.107_3
py27-setuptools: 40.8.0_1
...
Installed packages to be UPGRADED:
pango: 1.38.0_1 -> 1.42.4_1
libXrender: 0.9.9 -> 0.9.10_2
...
Number of packages to be removed: 115 Number of packages to be installed: 165 Number of packages to be upgraded: 521 Number of packages to be reinstalled: 104
The process will require 2 GiB more space.
2 GiB to be downloaded.
Proceed with this action? [Y/n]: 9.56 real 2.67 user 0.03 sys
I've seen problems with pkg upgrade before, but this is ridiculous. I had carefully
upgraded packages only a month ago, before my first system upgrade attempt. Why this
ridiculous number of changes? Something else must have gone wrong. I'm reminded of
Aussie Broadband's abortive attempt
at a network upgrade a few weeks
ago.
OK, revert to the old system again, not helped by the fact that I had thrown a switch
in /etc/rc.conf that changes its configuration:
# Decide whether this is a build system or a "for real" system.
# This is set to 0 during build, and 1 during normal operation
FORREAL=0 if [ $FORREAL -eq 0 ]; then
hostname="stable.lemis.com"
hostip=192.109.197.192
defaultrouter="192.109.197.137"
else
hostname="eureka.lemis.com"
hostip=192.109.197.137
fi
OK, fix that, reboot, and I was back where I was yesterday afternoon. The weather
observations database shows the real outage: last record yesterday at 15:19:26, first record
today at 09:57:43. Total outage a little under 19 hours.
OK, now lick my wounds. What went wrong? First, which X drivers did I have?
From /var/log/Xorg.0.log, the old system has:
[ 59.843] (II) NVIDIA GLX Module 304.131 Sun Nov 8 22:04:30 PST 2015
And the new system has:
[ 3065.285] (II) NVIDIA dlloader X Driver 304.137 Thu Sep 14 13:47:42 PDT 2017
The same driver! Why didn't it find anything? A bit of searching brought me to this page which,
after selecting “supported products”, confirms that version 304 supports both cards. So
it's just a documentation error: the first page just shows the most recent driver that
supports a card.
So: what went wrong? It seems that my configuration was correct, and yes, I had the module
loaded. More head-scratching. Sometimes I wonder if I made the right choice in buying
nvidia cards. My one recent experience with Radeon was surprisingly easy,
and it didn't require any driver downloads. Maybe it's time to investigate them more
closely.
Back to other issues. What was wrong with my ipfw firewall configuration?
A diff between the old and the new configuration shows a regression:
--- rc.ipfw 2018-12-31 18:36:05.000000000 +1100
+++ /destdir/etc/rc.ipfw 2017-06-07 13:33:42.000000000 +1000
@@ -1,15 +1,13 @@
# This should work
# GW is the external interface
-GW=180.150.55.199
-GW=167.179.139.35
+GW=180.150.113.90
WWW=208.86.226.86
natd -p 8668 -n xl0
ipfw -f flush
ipfw add 00010 allow ip from 192.168.20.0/24 to any
# external trusted IPs
ipfw add 00020 allow ip from $GW to any -# ipfw add 00020 allow ip from $GW1 to any
ipfw add 00021 allow ip from $WWW to 192.109.197.137
ipfw add 00022 allow ip from 127.0.0.1 to any
ipfw add 00023 allow ip from any to 127.0.0.1
I had a specific rule to bypass setup from the gateway address, but that had changed some
time in the past, and I hadn't updated it on the new machine. Another of those irritating
upgrade issues. I've been trying to find a comfortable upgrade path now for something like
a quarter of a century, and I don't seem to be much closer.
One thing that did clear itself up: why did the photo backup yesterday take so long?
Because I had deleted a whole lot of work files on /Photos since last time, and the
backup had to get rid of them. In fact, the average transfer rate was 33 MB/s, much higher
than the normal 10 MB/s or so:
Sat 13 Apr 2019 20:49:08 AEST Disk 9
Filesystem 1048576-blocks Used Avail Capacity iused ifree %iused Mounted on
/dev/da3p1 7,629,565 5,494,242 2,059,027 73% 1,806,255 1,437,391 56% /photobackup
sent 143,929,656,819 bytes received 1,069,707 bytes 33,129,411.10 bytes/sec
/dev/da3p1 7,629,565 5,418,704 2,134,564 72% 1,849,380 1,394,266 57% /photobackup
Sat 13 Apr 2019 22:42:10 AEST Photo backup ended
Call from Tomas Kucera of Effective
Electrical first thing this morning, on my mobile phone; I had asked him not to, but
he didn't seem to understand, and claimed to have called the same number that he called me
on on Friday. He wanted to
know whether to continue with the installation on Wednesday or not, maybe without the
batteries.
No way. I want those batteries, preferably both of them. He talked to somebody behind him
(his boss?), and I suggested that they work things out between them and that he then call me
back.
What does that mean? Maybe the whole issue wasn't so much the solar energy rebate as his
supply issues. I'm not happy about having to wait until October, and if he can't deliver
now, what are the chances of getting the second battery then? Time to investigate others;
at least I now know what I'm looking for, so it's not quite so painful.
First tried EverSolar, who had installed a
system for ibiza, a nick on IRC without a name. He was happy with the
installation, but EverSolar doesn't venture out into the open countryside, so I was outside
their installation range. Adelaide
yes, Brisbane yes, but
not Dereel.
Filled out some web forms and later got a call from Jim of AriseSolar, who accordingly calls himself Jim
AriseSolar <jim@arisesolar.com.au> later
on. Sydney number, not quite the local
presence I was looking for. But he came up pretty quickly with a competitive quote: $19,220
with 10.8 kW panels and (only) 9.8 kWh batteries. It's not clear why there was this
limitation, but he also planned 2 5 kW inverters, which seems sub-optimal to me.
Later a call from Tomas, on my mobile phone, but I didn't get there before it ended.
Couldn't be bothered to call back, and sent him a mail message instead expressing my
concerns. No reply.
Middle of autumn today, time for my monthly garden flowers photos. Gradually the weather is getting milder, though it's still
relatively warm. Today the temperature reached 28°.
The Hebes are a mixed bunch. The
ones I was worried about are almost certainly dead, but some of the others have started
flowering again.
I think they're too sensitive to drying out. Next year they'll all be planted in beds.
Since being freed from their big brothers, the
small Carpobrotus seem to be feeling
a lot better. This one in particular, opposite the house entrance, has developed a number
of flowers, though it's not really due until November:
The Hibiscus bushes are all still alive,
though their states are different. The “Uncle
Max” Hibiscus
rosa-sinensis looks happy enough, but it's been a while since it has a flower:
The Hibiscus syriacus are
similar: one (the one that Mick trimmed with a whipper-snipper) has grown happily over the
year, but has no buds, while the other has developed multiple flowers:
Unfortunately, the other plants in the north beds aren't doing as well.
My Salvia microphylla only has
a couple of flowers, and the blue ones are dead—for this year, anyway:
And the mistreated Clematis “General
Sikorski” (last month I had assumed that it was dead) seems also to be on the road to
recovery. Here on 11 March and now:
An interesting review
summary about the Canon EOS RP
from DPReview today, brought out after the
real review.
I'm not that interested in the camera, and the real review is too detailed for me, but the
title of this one (“Is the Canon EOS RP right for you?”) sounded interesting.
And it was. Apart from the lack of useful lenses and the abysmally slow frame rate (only 4
frames per second! Even my old Olympus E-PM1 does better
than that!), the really interesting thing was:
On the downside, if you like to mix stills and video in your travels, the RP won't be of
much help for the latter – if you have a reasonably modern smartphone, chances are its 4K
video will be leagues ahead of what the Canon is capable of in most lighting conditions.
So what good is it? I wonder when Canon will come out with a competitive mirrorless camera.
I've had a surprising number of issues with mobile phone batteries lately. Two different
old Samsung phones needed new batteries. In particular, Petra Gietz' I9210 died a couple of months ago, though the
battery wasn't that old. A replacement was dead on arrival. And then Petra
recently told me that the third was also dead.
There's something wrong there, and it's difficult to blame it on the batteries. Kept the
phone and put it on charge, and by elimination discovered that one of the USB charge
cords—and only one of them!—was defective. And the first of the three batteries wasn't so
dead that it couldn't be recharged. It does seem, though, that this particular phone is
very fussy about the charge current.
Only: I can't turn the bloody thing off! When I power it down and it's connected to the
charger, it automatically restarts. Without the charger, it works fine. That only started
happening with the old battery. Why? And how can I stop it?
Chris Bahlo is a typical modern person: her main form of communication is via Facebook, and she's pretty good at checking that mess at
frequent intervals.
So when Yvonne didn't get a response from her for some hours,
she got worried. Call her on the phone. Voice mail. No call back.
Chris lives on her own, and she has a number of animals to look after. What if something
has happened to her? Over to take a look, and found that she had left by car with Rev, one
of her stallions. Did she have problems on the way? Call up Amber Fitzpatrick and Cliff
Marisma? Sorry, don't have phone numbers.
Somehow this is all wrong. Facebook isn't the centre of the universe, and if you have a
mobile phone, you should be able to answer it, at least if you're not at home, and people
should have the numbers. What about the animals if she ends up in hospital?
Things turned out OK. She had been with Cliff (for whom we still don't have a phone
number), and had just been busy. But we should work out an emergency schedule for her
animals in case she's incapacitated.
Still no reply from Tomas Kucera of Effective Electrical by this morning, so I called him up to find out what had gone
wrong. He had tried to call me, he says, on my home number, the one stored in his phone.
No evidence of that at my end. Maybe the number's wrong? That's the one he called me on
on Friday. And why didn't
he leave a message? And why didn't he answer my mail message? As he said, he's a Microsoft
certified something.
Ah, he never got the mail message. The problem must be at my end. OK, here's my server
log... Oh:
Damn! Forgot to restart my mail tunnel after the last reboot. OK, one point to Tomas. But
he seemed to be excited because on Friday I had asked for time to think over what to do: he
had to schedule his sparkies, something that he hadn't mentioned earlier. And he couldn't
quote any other batteries because there are no other ones that would work with the
Ingeteam inverter. I told him about my discussion with Fyodor two weeks ago, but he didn't know any
of that information. Left him to discuss the matter with Fyodor and get back to me.
In any case, the good news is that he will, indeed, be able to deliver the first battery
pack tomorrow. And what happens if the second one isn't available on time? Dammit, this is
all too difficult, I have a "get out of jail free" card. Why don't you just take your
deposit back and we'll forget the whole thing?
Somehow he doesn't understand what a contract is. He referred to the terms and conditions
in his quote, but there were no such conditions in the contract that I signed. In any case,
that's not in anybody's interest. I asked to speak to his boss, but no, his boss isn't
interested in speaking with me—says Tomas. I'm sure that would change if we really had to
go to VCAT about the matter.
So we're going ahead with the installation tomorrow. Tony, the electrician, showed up and
did some more measurements, apparently related to the way the inverter needs to be wired (in
groups of three, it seems). We'll get 24 on this section of roof (facing 10°):
He also confirmed that Tomas can get “a bit excitable”, and made the relatively obvious
observation that if the battery coupling (basically just cabling) is not available on time,
it's relatively trivial to do some custom wiring. But he had heard from Fyodor that the
coupling is not just on time, it may be available earlier than expected. Preferably not
before 1 July, so that I can get the state battery rebate—maybe.
Message from Chris Bahlo (on Facebook, of
course) to tell us of two bushfires: one
in Mount Clear, and one in Grassy
Gully Road. The latter was supposed to be small.
No further information whatsoever. Earlier it has at least said that one vehicle had
attended the incident. But the map shows that it was about 500 m from hour house. Anything
to be seen? Nothing.
Later we went walking with the dogs, and of course looked for any signs of the fire. None.
But there was something there in the distance, a couple of hundred metres south of the
specified location:
Clearly nothing to do with the fire, but interesting because it shows so much more than we
could see with the naked eye. Something like a delayed action telescope.
Interesting local news on the (state-wide) radio news broadcast: apart from the bushfire in Mount Clear, there was a mention
of the World Jousting Tournament[dead link]
in Federation
Square, Melbourne.
Jousting? That's what Chris Bahlo does,
with her mates Cliff Marisma and Phillip Leitch. Today's news report gave two names: the
current world champion, Phillip Leitch, and his Australian contender Cliff Marisma. It
seems that nobody has twigged to the fact that Phil is also Australian (he runs the show
at Kryal Castle). And the page
above shows a particularly poor photo with the text “Reigning world champion Phillip Leitch
(red) and Australian team mate, Cliff Marisma of Australia perform a medieval [sic]
solid lance joust...”:
Dead image link: http://www1.pictures.zimbio.com/gi/Medieval+Jousting+Display+Melbourne+CBD+TW9nxvBtaXtl.jpg
Only three years later, these links are dead. There's plenty to be found about jousting
on Google, but I no longer know what the contents of the links and images above are,
and exceptionally I didn't keep a copy. Possibly this one could be related:
But I identify them more with paintings in Yvonne's office:
The electricians from Effective
Electrical along first thing in the morning with a surprisingly small pack of 36 solar
panels, and set to making a fiendish noise attaching clamps to the roofing and rails to the
clamps.
The wires seemed to be very thin. The label shows that the panels can deliver up to 9.72 A,
and there 12 of them connected together. But as Tony told me, they're in series, so that's
the total current. But the open circuit voltage of the panels is 40.1 V, so the array has a
voltage of 480 V! That's much more than I expected.
And by evening all but two of them had been mounted:
In the course of the day, Tony and Fyodor came along and brought in the battery and the
inverter. It seems that Tony was wrong about the “coupling” just being cables. It seems
that there's a lot of electronics involved, and that's why it has taken so long. But
they're doing a field test soon, so I asked Fyodor to put me down as a tester. That could
still happen.
With all the things going on in the front of the house, let the dogs out to have a look.
But somebody had left the gate open, and despite the interesting stuff going on in the
driveway, Leonid decided that Grassy Gully
Road was more interesting, and set off by himself. By the time I caught up with him he was
half way along. I really hadn't expected that.
Last week I discovered that
my “new” GPS navigator (less than 5 years old, anyway) had had
a Y2K event and reset its date to August
1999. And that made me think: what would my ancient Garmin GPS2 GPS receiver do?
Went looking, but couldn't find it. But coincidentally, because of the solar panel
installation, I had to move the fridges out of the garage and into the shed. OK, take a
look in the old cartons that have been there since we moved in... GPS receiver, a digital
level and a whole lot of old important CDs that, for some reason, I haven't needed in 4
years.
OK, fire up the GPS 2. It works, though a row on the LCD screen has died. And it gave me
the correct time! It must have been through two W1K rollovers. How did it manage?
I've forgotten who gave it to me, but it must have been well over 10 years ago. I had been
told that it was a Peruvian lily, but that seems to
be Alstroemeria. And Laurel Gordon
told us that it was
a Mirabilis jalapa. But I had
already noticed discrepancies: Mirabilis cross-breeds, while our yellow plant remains yellow
and the red one remains red. And they call it the “four o'clock flower” because it doesn't
flower until 16:00. These photos were taken at 7:30. And finally Mirabilis is a bush, and
these are tubers. So what are they?
Fyodor was there again supervising the installation of the equipment. He later took some
photos with his phone, and told me that this was of great interest because it was one of the
first installations of this size and in this area.
It's surprising how much wiring is needed for these systems. They told me that they had to
install a number of oscillators, but I later discovered that they meant isolators—it's clear
where I'm coming from (and, interestingly, where my spelling checker comes from: it flags
“isolator” as a spelling error). But what a number of isolators! One at the end of each
row of panels (the orange ones) and a whole lot more in the garage:
The one on the right of the inverter (second photo) is double-throw Turning it off isolates
the inverter and feeds grid power to the house. The ones on the top left are for the three
individual PV arrays, and the one underneath is for the batteries. The pipe on the right is
part of the vacuum cleaner system, nothing to do with the wiring.
Round 14:00 they rewired the main switchboard to go from main switch (left) to the “Inverter
in” switch (second from right), then back from the inverter to the remainder of the
switchboard (far right). Hopefully for the last time we needed the UPS to bridge the 10
minutes that that took.
They were done round 15:00, and Fyodor showed me how to access the configuration from a web
browser. Lots of nice knobs to turn. As discussed, he had set the thing to stop using
battery when it dropped to 20%, as long as the grid was connected; otherwise it would
continue using it down to 5%, the minimum practicable (and not 0%, as Tomas had claimed).
He had also set it to charge from the grid at up to 500 W when it reached 20%. It's
not clear that that's a good idea, but hey, I can change it if I don't like it.
Spent some time setting up an account with Ingecon, with some rather interesting details. Clearly you need to enter a device ID
and password, and clearly you need to select “Verify connection”. But this photo
is after successful verification!
This was done the following morning. I think that it purports to show solar panel input on
the left (2335 W), consumption in the house (7306 W), power from the grid (4971 W) and power
charging the battery (2838 W). How does this add up? And why is the battery being charged
from the grid? Fyodor put a 500 W cap on that, and as I discovered during the evening,
after dropping to 20%, it charged back up to 30% from the grid. Does that make sense? I
need to think about it.
In addition, the somewhat meaningless usage statistics are stated for Saturday and Sunday,
without any date. Does the inverter not know the time? It has a configuration entry
for NTP, which I set. It should know that
the days are Thursday and Friday. And sure enough, it shows the time correctly, in emetic
US American date format: 04/18/2019. The only other language that the device does
is Spanish, but maybe that's preferable to that horrible date format.
Three months ago CJ Ellis
and I tried to mount a rod between the wall and pillar in the house entrance. We failed:
the bricks were too hard. So today Yvonne asked the
apprentice sparky if he could finish the holes. Yes, in record time, with a bit of advice
from his foreman. It was so fast that I didn't have time to get a photo of him.
Was that good luck or good equipment? I had started the first hole in January, and he
finished it in a second or two, breaking through to a hole in the brick. Might I also have
done that? And was the second one softer, maybe? I had had one like that on the other
side. Still, it's done.
Yes, nicely bent out of shape. It seems that both transistors had been destroyed, and I'd
guess that more are as well. But doesn't it make things easier to have well-graded photos?
Getting hold of Daniel O'Connor's photos wasn't as easy as it sounds. In principle if you
have an image in a web browser, you should be able to save it. But Daniel's photo is
https://www.icloud.com/photos/#04YWjMDeSreIQ-Fi7InYEGESw. Apple.
No, you can't just save it, you have to ask nicely by signing in with your Apple ID. OK, I
can do that, since I couldn't find a way to delete it.
Not so fast. First we'll send you an SMS to * ******70. Ah, that must be my normal phone
that is always on auto-answer. Forget it.
Once again, eureka can't keep quiet. Blank the screen and it's back on again within
seconds. What's causing that? I had had various programs under suspicion, but this evening
I couldn't be bothered. Switch to X server 1, which
doesn't have anything unusual running. But that didn't stop it.
So what is causing this? I can't see any way that a program connected to one server can
want to influence another server; there are few enough computers that run more than one
X server. More head-scratching.
After some comparison, it seems that the “PV Generation” and “Grid Consumption” figures add
up to the number in the middle with the house symbol. And the “battery charge” figure below
is not directly related; to get the house consumption you need to subtract that number
(7306 - 2838 = 4468 W). The dates are ridiculously wrong. It's
even counting in the wrong direction: yesterday was “Saturday”, but now it's “Sunday”, and
today is “Saturday”:
And there are no useful statistics. How much PV power did I generate today? How much did I
feed into the grid? How much did I take out of the grid? Nothing there, just meaningless
comparisons to some arbitrary concepts.
“Daily consumption of 1 home”. What kind of home? “0.00 barrels of oil”, saved by not
drawing current from a coal-fired power plant. “4.07 kg of CO2 [sic]”. How was that
calculated?
Instead I got a report from Spain with marginally more information:
“Consumption 15.02” (we're too leet to specify units; everybody knows that we mean kWh).
What does that mean? The other second explains: “From PV: 15.02”. So that's what we used
directly? And then there's “Battery Charge” and “Grid Feed-in”. At first I thought that
they matched “From storage” and “From public grid”, but no, I think the first are what we
transferred from PV to these destinations, and the second is where the power came from that
we used. But that's all still pretty much bare-bones.
But wait, there's more! Go to your web site in Spain (this is clearly too important to keep
on your site) and see:
Potentially that's more useful. It's interesting to see the tooth-like pattern during the
night where the battery alternately charges and discharges. And round 7:00 the big spike is
turning on the air conditioner to warm the place up. The next one is presumably cooking
breakfast. I need to analyse this stuff (and find a way to fix the broken text rendering),
but it could give insights.
But is it correct? The day before shows a peak usage of somewhere between their chosen grid
values of 3.7 and 4.6 kW:
Round 19:00 I noted a consumption of over 7 kW. Who is lying? It certainly could be the
inverter, which shows a lot of to-and-froing between exporting to the grid and importing
from the grid in a time frame of 5 seconds or so. Is it really doing that, or just
reporting it incorrectly? The display on the inverter itself shows the same thing: here I'm
running off the PV array and charging the battery (at 7.7 A), but it claims to be feeding
power to the grid as well (admittedly only 11 W):
About the only thing that's clear is that I didn't get the promised alarm that the system
was off the grid.
Still, what I'm seeing here is a total power generation round 31 kWh for today. Given that
we're in the winter semester, and that I had expected the equivalent of about 3.5 hours
“sunshine” on average for the year (37.8 kWh) that's not too bad. But of course today was
sunny; we'll need to factor in the cloudy days.
Still, I want to tweak the system. Where's the configuration? It seems to be hidden where
I can't get at it. I have a password for the inverter, but it doesn't seem to apply to the
web interface. The documentation talks of a “password” (« contraseña ») 0332, which
I would call
a PIN. I need to
tread carefully here; there's every possibility of locking myself out, and I don't want to
do that.
Yesterday's investigations
brought me to two partial conclusions: the web application on the inverter has serious bugs,
and I don't have access to the configuration menus. I could continue trying to break in,
but it's probably a good idea not to do anything until the people come to inspect it. That
should happen this coming week, but considering that Monday and Thursday are public
holidays, I'm not betting on it.
So what can I do? Clearly there's a lot of information going out of the inverter. How
about firing up a wireshark and seeing
what's going on? First stop all the web browser sessions and then start one and watch what
happens.
First problem: after stopping them all, there was still traffic. Where was it coming from?
I couldn't find out, and in the end I had to run the capture to ignore its port number. And
then, sure enough, there was plenty of stuff to see. The most interesting URLs were
http://inverter.lemis.com/ems/summary and http://inverter.lemis.com/ems/last7Days. OK, I can get that too. The
information looks like:
That's interesting, but it doesn't make much sense. Why is the StartDate the way it is? On
11 March 2019 I didn't even know about the Ingeteam offerings. And why is there no time? I've
already established that the inverter has NTP
and had it set correctly.
And the other values? The identifiers don't make much sense. And it doesn't seem to be
updated very often. At first I thought it might be a web cache thing, but that shouldn't
happen for local sites, and adding the -n option to fetch didn't change
anything.
And last7Days? There's more information there, multiple records, probably the
background for this broken display:
At least the date is correct (though in a different format from the summary record),
so there's really no excuse for the broken bar chart. But it's still not clear what the
numbers mean. More head-scratching needed.
Yesterday's network sniffing
gave some insight into how the photovoltaic inverter communicates, but not enough. Spent
some more time sniffing with ngrep and
found:
Ah! That's the real data, not the locations I found yesterday. And this one does
have time stamps that make sense. OK, take a look at http://inverter.lemis.com/ems/sse/stream.data:
=== grog@eureka (/dev/pts/29) ~ 1 -> fetch http://inverter.lemis.com/ems/sse/stream.data fetch: http://inverter.lemis.com/ems/sse/stream.data: Not Found
Hmm. What's this stream stuff? Must be something that I don't understand. But there's an
alternative, the dirty one that Daniel O'Connor suggested: sniff the data stream and collect
what I'm looking for. Here I have clearly identified source and destination IP addresses,
source port, and the beginning of the packet data.
So how do I do that? BPF, of course, and more specifically pcap. Is there a program that will do
this for me? You'd think that ngrep
would do it for me, but it wasn't obvious from the man page.
Further searching brought me to this page, a tutorial with
source code. Played around with that for a bit. The first program, ldev (source in
the text), compiled out of the box and ran:
pcap_t *pcap_open_live(char *device,int snaplen, int prmisc,int to_ms,
char *ebuf)
snaplen - maximum size of packets to capture in bytes
promisc - set card in promiscuous mode?
to_ms - time to wait for packets in miliseconds before read
times out
errbuf - if something happens, place error string here
*/
Where do I go from there? The best guess from the code is that it doesn't like -1
as a timeout value. But do I really want to get involved in debugging? I just want a
filter. Maybe I should learn about streaming after all.
Yvonne told me what I should have known today: time to clean
the bore water filter. I had decided that it needed cleaning every week rather than every
few months, probably as a result of the low water level due to the dry summer. But I had
neglected it for a couple of weeks, and it was very much in need of it.
That wasn't the only issue. It's been weeks (months?) since I decided I need to look at
circuit 4, which has only too clearly been leaking:
We've been planning the solar electricity for well over 4 years, and much of the delay was
due to procrastination on my part, coupled with a complete inability to establish contact
with Simon Reid of BREAZE. Chris Bahlo
didn't have that luxury: she has no grid connection, so she had to have it installed for her
new house (also four years ago).
Today she came along and we compared notes. She paid about the same amount of money as we
did. For that she got 7 (count them, seven) solar panels and a lead-acid battery of
unspecified dimensions. But since she's only generating 1.75 kW, and the average hours of
sunshine here are round 3.5 hours, it's unlikely that the batteries would store more than
7.5 kWh. And of course her inverter is completely dumb.
That's quite a difference, though even at the time I thought that the price was excessive.
Maybe it wasn't such a bad thing after all that I didn't go with BREAZE.
Easter Day today, traditional time for
eating lamb. And we have had a boned leg of lamb in the deep freezer since last Easter
(Why? We ate lamb then too. Did we buy two of them?).
As mentioned at the time, the thing was so big that I cut it in two.
OK, thaw it out and roast it.
Not a success. I've been trying to work out roasting times for meat, including lamb, for
some time. But one thing I really must do is ensure that the start temperature is
the same each time. It should be room temperature (say 22°). If you take it straight out
of the fridge (say 4°), like I did today, and you want to cook to 60°, you have a
temperature difference of 56° instead of 38°. That's nearly 50% more, and it makes a big
difference in the result. In fact I took it out when the thermometer was showing 52°, and
it was still marginally overdone.
Yesterday I investigated
sniffing the traffic between the inverter and a web browser, in the assumption that it would
be easier than recreating the requests needed for communicating directly with the inverter.
But Jamie Fraser and Daniel O'Connor (on IRC) saw it differently. Jamie suggested using the
developer tools (F12 on most browsers, it seems). I've seen them before, but never used
them in earnest, especially not to analyse HTTP streams. OK, try it out:
That's not immediately obvious, but none of the traffic I'm looking for appears there. Ah,
says Daniel, works on Chrome and Safari. OK, find a Chrome and try again. Great
improvement. Now I can see the data coming in, and as expected without any GET requests:
That looks a lot better. In particular, the progress bar at the right shows that it's
continuous. And sure enough, select it as a curl command, and
we're away!
What next? Pipe the data into a database, of course. Daniel has been there before as well,
and he has an application, written
in python, which
collects the data, puts it in a database and even generates graphics. Took a look at it and
was discouraged: it's relatively long, and my Python-fu is not really up to it. There must
be a simpler, if kludgier, way.
I'm sure I've seen this one before, but it's worth identifying: it's a time of year when not
much is flowering. Maybe I should find a few autumn-flowering bulbs.
The real surprise is that it doesn't happen more often.
Why are Australian doors of such poor quality? They don't shut properly, the joiners
install them with a 3 cm gap at the bottom, and they have these positively dangerous
fittings, though I suspect that JG King have chosen a particularly poor example.
More playing around with the solar power data stream today, and managed a partial conversion
to an SQL INSERT statement, complicated by my lack of recollection of the syntax.
But I'm getting there. The question remains whether I should do a proper parse of
the JSON data, but I think I can get by
without it. Other questions are more interesting: what do I do with the data? There's one
record like this every second:
All those numbers appear to fit into 16 bits, but there are quite a few of them. 86,400 of
those a day? I have a comparison with my weather database, which has been collecting one
record a minute for the last 9 years. I now have 3,800,000 records, and
the observations table is 280 MB in size. I can live with that, but with one record
a second I would reach that size in a couple of months. For the time being I suppose I
should just collect them, but probably I'm going to have to remove records older than a week
or so.
And what do the fields mean? I think I have cracked most of them. These ones appear in the
once-per-second records:
This interpretation is only partially correct. Don't use it.
Identifier
Meaning
Units
Comment
Id
?
IP address?
Pac
"Power AC"
W
Current house consumption
SetPoint
?
numeric
Always 0
Alarms
String
Status
Inverter status
String
Only ever seen "On-grid"
SOC
"State of Charge", battery
%
VBat
Battery voltage
V
PacGrid
Power from grid
W
Does not show power to grid
PacBat
Power from battery
W
- means charging
PacPV
Power from PV array
W
FromPV
?
Boolean
Always false, even when drawing power from PV
Codes
?
Never seen any set
W1
Power to grid
W
One of three (possible) phases
W2
Power to grid
W
One of three (possible) phases, always 0
W3
Power to grid
W
One of three (possible) phases, always 0
And these appear about once a minute:
Identifier
Meaning
Units
Comment
Id
?
IP address?
As before
Date
Date and time
String
Format "2019-04-22 13:57:00"
Phase
Phase number on grid
numeric
Always 3 of these, even on a single phase connect
W
Grid power
W
Negative → grid
Pac
"Power AC"
W
PV power input, not current house consumption!
PacDischarge
?
?
Always 0
PacCharge
?
numeric
Digit between 0 and 3
PacPVBatt
?
numeric
Appears to be close to Pac (power from PV array)
EMS_VBAT
Battery voltage
V
VBat in other record
EMS_SOC
State of charge
%
SOC in other record
WatDigitalInput
?
numeric
PacRev
?
numeric
HWOutputs
?
numeric
Clearly there are a number of contradictions here. Different programmers, or inadequate
planning? And does the thing really always output for three phases, or is this a matter of
the inverter configuration?
I'm quite happy with the amount of power we're saving even at this time of year; the daily
reports in triplicate from Spain show that I'm generating between 53% and 66% of my total
usage. But it's worth thinking about how to ideally use the power. The algorithm used by
the inverter is:
Use power from the PV array first.
If the PV array can't supply enough power, supplement from the battery.
If the battery can't supply enough power, supplement from the grid.
And then there's the other view:
If the PV array can supply more power than needed, charge the battery.
If the battery is charged, feed the power to the grid.
All this appears to make perfect sense. But already there's a first problem: what if the
battery is charged and the grid is being supplied at full rate, and there's still a reserve
of energy? It gets lost. And from my observations, that happens quite a bit of the time.
The second battery, when it comes, will reduce that time, but not eliminate it. So clearly
there are things that should be done when the sun is shining, like running dishwashers and
washing machines.
And then there's another situation that is even more difficult to understand: should I
charge the battery first and only then feed to the grid? Today the battery started charging
round 8:00 and was done by 12:00. Then I fed to the grid for a total of about 3 hours, in
the process running into the limits. If I had started feeding in to the grid earlier, I
would have been able to feed in more, possibly double the quantity, and still charge the
battery.
The problem is: I know that now. How could I have known it this morning?
Another attempt at “air-fried” Bratwurst today. Based on previous experience, decided
to “fry” them hotter and longer, 210° and 15 minutes instead of the 200° for 12 minutes last
time. The higher temperature seems to have been a good idea, but it seems that the original
time was sufficient. Here before and after:
I've spent the best part of a week thinking about how to process data from my solar
inverter, and I think I'm on top of how to get it into a database. But that's only part of
the story. What do I do with the data in the database? Do I want to create another page
like the weather page? In principle the pretty
displays that come with the inverter do most of that. So maybe it's time to step back and
do something else.
One thing that strikes me: what is using all the power? The daily reports (the ones that
come in triplicate) confirm what
