First day of summer today, but you'd have to know it. The maximum temperature was 13.3° in
a month where temperatures regularly exceed 40°. As the ABC news put
it:
Phở bò for breakfast today. I have two different
jars of paste from Por Kwan, and for the fun of it I chose the other one. I thought. It
seems that they make no less than three different pastes, all with labels that don't say
very much in English:
“Vietnamese Beef Flavor Paste, “Instant Beef Flavour Paste”, “Spicy Beef
Flavour Paste”. Not much difference in English, since they're all Vietnamese, “instant” and
“spicy”. The one I got was the third one. Not much help from Google Translate:
Finally braved RIPE's login
procedures to set up a reverse domain for 192.109.197.0/24. The good news: login
worked. The bad news: it was only the first step. I received a mail
Two-factor authentication (2FA) is mandatory on RIPE NCC Access. To set up 2FA on your
RIPE NCC Access account, please copy and paste the below link in your browser.
More information on how to set up two-factor authentication can be found in the link
below.
Following the link gives me two choices: one of three Authenticatorapplicationapp or a passkey. I had never heard of either, and the instructions are vague: “Then, either
scan the QR-code on the set up screen or enter the secret key manually”. Wha setup screen?
How do I scan it? I need a mobile telephone for security? But it seems
that's becoming more and more a requirement: if you don't have a mobile phone, you don't
exist.
Spent much of the afternoon trying to understand how this fits together. It's worth taking
a step back and considering what they're trying to achieve. For a horrible moment I feared
that they would want me to perform the operations on my phone instead of a Real Computer.
But that doesn't make sense: for any reasonable 2FA you need at least
two devices. I have already ranted about the stupidity of performing 2FA on the same device
(typically phone) from which the request came. But they don't know my phone number, and
even if they did, would they really send me a message round the world? How do they
communicate with the device? I could use hirse.lemis.com, my old Xiaomi Redmi 9T that no longer
works with Australian mobile phone networks, but only if it uses IP (as one would expect from
RIPE).
So what kind of authentication makes sense? What I see is a six digit code. Is that more
secure than a random password? What does make sense is something like face or
fingerprint authentication. That can also be circuvented, but it's an order of magnitude
more secure than silly things like dates of birth.
The real issue is just understanding what they want, especially since they say that once
chosen, the method can't be changed. What if it doesn't work? They allow only three
authenticators: Google
Authenticator, Microsoft and FreeOTP (which they spell as “FREEOTP”).
Then there are the passkeys. OK, a number of sites accept them, so I can use one of them in
case something goes wrong. Spent some time trying to set it up on my computer for
Amazon, but failed. At
one point I got this message:
I'm getting to hate these silly values with lots of *** in them. OK, enter the (6
digit) number. It didn't like it for some reason, and it took me several attempts and lots
of deliberate delays to finally authenticate. But then I got:
What's the Security key? I know the Any key. Have they added another key to the Microsoft
keyboard? Much searching shows that it's a piece of hardware such as the dongles following
FIDO2. And that
makes sense: they're one thing that mobile phones have that Real Computers don't. I need to
follow up on that. What software support do they need? Will they even work on a FreeBSD machine?
It looks like lack of fertilizer, though I have spread some not so long ago. Try again with
much more fertilizer; hopefully new growth will look better.
The batteries in my bathroom scales are low. On changing them, I discovered that they were
nickel-zinc batteries, and only one was low. That's typical for NiZn batteries, and one of the reasons
I no longer use them much. But on replacing them with normal NiMH batteries, the display was too dim, probably the reason why I put in the NiZn batteries.
Apart from that, the result wasn't what I expected. I had just weighed myself with clothes
and registered 90.8 kg. But now it showed 92.6! A few months back I had similar problems with brand new scales, but this time it was
diffeent. A comparison with Yvonne's scales showed that the
90.8 kg were correct. But when I weighed myself again on my scales, they too showed 90.8
kg.
OK, it seems that the scales don't like the voltage of the NiMH batteries. Replace them
with NiZn. 93.4 kg! A bit later they had reconsidered, and were now showing 91.3 kg. And
much later they again showed 90.8 kg.
What's going on here? Do the scales somehow self-calibrate? It's certainly different from
my last investigations, where the scales were consistently wrong.
Bruno has been transformed by being let
out when he wants, and I think he's still finding out when he wants to be inside and when he
wants to be outside. But to be on the safe side we don't allow him into the Great Outdoors
after sunset. Instead we put him in the “dog run” outside Yvonne's bedroom, where he can't get out.
Until tonight. While watching TV, he appeared on the window ledge and wanted to get in.
And he had a dead superb
fairywren in his mouth! How did that happen?
It proved that he dug under the fence in the dog run. But what was the fairywren doing out
alone at night? Still, it was his first bird in nearly 10 months. Hopefully they'll keep away from him from now on.
Where are all these failure messages coming from? On the face of it it could be a bug in my
PHP code. But the only plausible
place for a bug would be in responding to a 404 error, and they didn't happen. The best I
can do is to just not report errors that have a referrer http://www.lemis.com.
But today I got a whole lot of other errors:
1304 N + 03-12-2025 To groggyhimself@ World Wide Web Owner ( 78) N + Broken link: /grog/Photos/20160213/small/Radioactive.png <- http://www.lemis.com/grog/diary-feb2016.php?dirdate=20160205
1305 N * 03-12-2025 To groggyhimself@ World Wide Web Owner ( 78) N * Broken link: /grog/Photos/20160213/small/Confusion-symbol.png <- http://www.lemis.com/grog/diary-feb2016.php?dirdate=20160205
1306 N * 03-12-2025 To groggyhimself@ World Wide Web Owner ( 78) N * Broken link: /grog/Photos/20160213/small/Mining-symbol.png <- http://www.lemis.com/grog/diary-feb2016.php?dirdate=20160205
1307 N * 03-12-2025 To groggyhimself@ World Wide Web Owner ( 78) N * Broken link: /grog/Photos/20160213/small/Information-symbol.png <- http://www.lemis.com/grog/diary-feb2016.php?dirdate=20160205
1308 N * 03-12-2025 To groggyhimself@ World Wide Web Owner ( 78) N * Broken link: /grog/Photos/20160213/small/Extruder-symbol.png <- http://www.lemis.com/grog/diary-feb2016.php?dirdate=20160205
...
recognize fire symbols. Then there is confusion (represented by the <?php oneimage
("/grog/Photos/20160213/small/Confusion-symbol.png", "confusion symbol", 40, 42, 1); ?>
symbol), radioactive waste (represented by the <?php oneimage
("/grog/Photos/20160213/small/Radioactive.png", "radioactive waste symbol", 50, 48, 1); ?>),
mining (represented by the <?php oneimage ("/grog/Photos/20160213/small/Mining-symbol.png",
"mining symbol", 45, 44, 1); ?>), information (represented by the <?php oneimage
("/grog/Photos/20160213/small/Information-symbol.png", "information symbol", 50, 41, 1); ?>
symbol) and extruders (represented by the <?php oneimage
("/grog/Photos/20160213/small/Extruder-symbol.png", "extruder symbol", 44, 41, 1); ?>
symbol). What do they really mean? Running the cursor over them gives you insight: the web
...
Oh. When I moved my images to DigitalOcean I had to rewrite my already convoluted showphoto ()
function to point to DigitalOcean when the
subdirectories Photos/*/big/, Photos/*/small/ or Photos/*/tiny/ were
referenced on the external web sites. But I didn't do it for oneimage ().
More pain!
In fact, it was simpler than it should have been. Just
prepend https://lemis.nyc3.digitaloceanspaces.com/ to the URL for local images.
Looking back, that shouldn't have worked: it should already have
had http://www.lemis.com/ there. But so far it seems to have worked, though I went
through a lot of pain relearning PHP in the process.
I've made phat si-io a number of times, but
it's still not perfect. This morning, since I still have a cabbage, I tried with that
instead of gai lan, and also
tried a number of different changes.
The problem: wok hei. That's a
flavour nuance that I haven't quite understood, but it's related to cooking over high heat
in a wok. Google Translate is completely useless. It translates the term into simplified
Chinese as 我开黑 (I play with friends), or into traditional Chinese as 我看黑 (I see
black). Wikipedia wrote it
“wohkhei”, and claims that it is a Cantonese word meaning
“the breath of the wok”. It also gives the spellings 镬气 (simplified), which Google
Translate translates as “gas", and 鑊氣 (traditional), which it doesn't translate. And then
MARION has a page on
the subject.
But this is Thai. Google Translate tells me that it is กระทะ เฮ้ (kratha ḥê). Is it right?
At least it translates it the same way from simplified and traditional Chinese. In English,
I get “gas” or “wok hei”, neither very useful.
So how do I get my กระทะ เฮ้, when I don't even know what it is? My best bet is that it's
something that's slightly charred, thus the continual recommendations to cook the noodles
very hot and by themselves. But you can char cabbage too, so that's what I did:
And then there are the noodles. I freeze them and then thaw them when I need them, and
they're a little moist. So I left them to dry out for 40 minutes before frying:
But they still stuck. There's something basically wrong in the recipes that I have: fry the
noodles with the sauce in a very hot wok. But that doesn't work: “very hot” should mean
200° or more, but the sauce boils at round 110°. So I fried the noodles in oil first and
then added the sauce. They still stuck, and left a lot of residue:
That's probably because of the lack of green (gai lan) and yellow (egg yolks).
The other issue that is probably more interesting is that the steel pan in which I made them
is slightly convex, so any oil drains to the sides. That's the exact opposite of what a wok
would do. So next time I'll at least use a wok.
Bratwurst for dinner today.
Normally I fry them, but why? Despite the name („Brat“ could be taken to mean “fry”), they
should be grilled. Just what we want for the hair dryer.
OK, off to YouTube and looked for some clips,
coming up with this one. Slit the sausages like Laugenstangen to avoid having them burst, then grill for 10 minutes at 160°. Turn
and grill further for another 3 minutes at 180°. And while you're at it, grill some
potatoes in a second “air fryer” for 10 minutes at 180°.
OK, we can do that. Slit and grill for 10 minutes. Here before and after:
What a lot of work! And of course I should have read my diary, which told me 14 minutes at 210°. And apart from that, I now have two dirty
“air fryers”, so that'll be another dishwasher full.
Like Bruno, Mona typically comes into the house via a specific
lounge room window. But today she looked different: something sticking out of her mouth.
On closer examination it proved to be the bottom half of a lizard. Off to get my camera,
but when I got back there was only this to be seen:
Yesterday the temperature reached 34.1°, and today it was 32.1° (or 26°, as the BoM predicted). Gradually it's beginning to look like
spring.
Strangely, some things are looking better than I recall. The Anigozanthos that Diane gave us
over 10 years ago has struggled for a while, but it's now looking as good as I
recall:
I know that (but not its name). It's a rather pretty weed, with flowers about 1 cm across.
But this is the first time that I have seen so many in one place.
And somehow the white irises are everywhere, even in front of the front door:
The last of the Cyber week special offers expired today, notably PhotoWorks from AMS Software. I've given up on trying to denoise
photos, but what about background removal? But before downloading Yet Another software
package, how about checking my existing packages, notably Gemstone 12. I got it free, but
it's from ACDSee, which I tested
last month without being impressed. Was that Gemstone 15? The product structure is
confusing enough that I can't be sure. I was negatively impressed both by the
performance and the user interface. Surprisingly, though, Gemstone 12 seemed almost normal.
It wanted to put me in the “folder” P:\2-grog, probably because I used it there before. But my current test photos
are in P:\1-Skylum (hysterical raisins, and no longer appropriate). OK, set the
directory and try to load a file. It still tried to load from P:\2-grog! OK,
set the “folder” there too, load the file. What I had from DxO PhotoLab (here original and DxO version)
was:
That's seriously suboptimal, and I started playing around first with the noise in the shadow
area. The results were very different from other software: it crashed, repeatedly. And the
first couple of times it forgot all the settings that I had made. I needed to stop
normally, and then it recalled the settings. Probably there's a way to save them without
stopping, but the program abounds with strange icons, and I didn't want to investigate yet.
A bit more playing round suggested that the recovery of the original image could be better
then DxO, but I didn't have a comparison. And then it occurred to me: before expending too
much time with last year's model, how about Affinity? I have the latest version
there, and they've promised free updates. But that will have to wait for another day.
And PhotoWorks? Not necessarily its forte. They claim “Intelligent photo editor with a
content-aware AI. Batch process multiple photos, boost colors and retouch portraits
automatically, remove unwanted objects, change background, etc.”. Is that enough? Not
today, anyway, so the last special offer is over and done with. I'll look again next year.
OK, I know what to do: load the images individually and join them one at a time. But it
didn't work! Firstly, it didn't find the images (specified on the command line). So I had
to climb trees to find the files. And when I loaded the first two and aligned them, I ended
up with this display:
OK, add my own. It didn't work! I couldn't position the cursor where I wanted it. It kept
moving somewhere else. Somehow the whole display is messed up.
OK, what happens with the old version on eureka? Off there and ran my script (which
runs pto_gen to generate the .pto file), and it worked perfectly. What's the
difference in output? I should take a look. The difference between the .pto files
may give me a clue.
And, of course, whose fault is this? Does this only affect the FreeBSD version, or is it a general
problem? My guess is that the not loading the files is general (“passing parameters on the
command line is evil"), but if other people had had so many problems with the fast panorama
preview, I think we
Jesse Walsh along today to do mainly weeding, but he also brought a “new” lattice to replace
the one that was falling apart (and which I forgot to photograph). Here's the new one:
Choucroute garni for dinner this evening.
I've been cooking it occasionally for over
23 years, and I continually find that the recipes have too much meat and too little
Sauerkraut. My first attempt
23 years ago had a ratio of 9 parts meat to 10 parts Sauerkraut. I had reduced
the quantities last year, and today I put in even less (1 part meat, 2 parts Sauerkraut), but
there's still too much meat. 40% maybe?
Is that really a result of the fertilizer? At first I thought only old leaves had fallen,
but this image comparison (with Monday) clearly shows that new growth is affected too:
Can an AI bot help? Google Gemini says
“no”, but points to iNaturalist and PictureThis. Well, no, it's too
polite to pointlink. I had to search for myself. And they're both mobile
phone applications! Well, iNaturalist is much more, but I had to install the app, take
the photo and then describe it on a glass keyboard. Not my thing.
And PictureThis? Full of spam when I started it, and when I finally got past that, it
wanted a credit card number. Another non-starter. About the most interesting thing
is part of its home page, which shows:
That's all that is shown. Why a text describing Frangipani (top right) along with a much
larger image of a Hibiscus rosa-sinensis? There may be a good reason, but on their home
page it's just plain misleading.
What about other bots? Claude. I couldn't move the image to its input field. OK, give it the URL of the
publicly accessible version. “Can't access file”. Why?
You're right to question that! Let me try fetching it again - sometimes the initial
attempt fails but a retry works...
I'm getting the same error - "Image content is not supported". This is a limitation of the
web_fetch tool I'm using, which appears to not support directly fetching image files from
URLs, even when they're publicly accessible. The tool is designed primarily for fetching
HTML pages and text content.
However, if you upload the image directly to our conversation using the attachment
feature, I'll be able to view and analyze it for you. That would let me tell you what's in
the photo, identify the plant if it's indeed a weed, and provide any other information
you're looking for.
How do I attach the image? Use the paper clip. No paper clip on my screen. At some point
it divulged that it could also be a + symbol. Aren't these interfaces inutitive? That worked, and I got some useful
information:
This appears to be Scarlet Pimpernel (Anagallis arvensis), also
sometimes called Poor Man's Weatherglass. The key identifying features I can see are:
Small, delicate 5-petaled flowers in a peachy-salmon/orange color
Purple or reddish center in each flower
The flowers appear to be about 1/4 to 1/2 inch across
Growing low to the ground among other vegetation
Small, opposite leaves visible on the stems
Scarlet Pimpernel is actually considered a weed in many areas, though it has quite pretty
flowers. It's called "weatherglass" because the flowers close up in cloudy weather or
before rain.
How about that! I had heard of Scarlet Pimpernel, of course, but
I didn't know that it was the name of a flower. And it's not even scarlet! And I had
thought that this would be some kind of Australian native.
VicRoads have sent me email. Not the
promised email confirming the registration of Yvonne's car,
but the information that they have a new and improved web site.
It certainly looks different. But it's still hard to navigate, and it has managed to forget
my password: I had to reset it. It looks as if passwords will go away soon anyway: now it
would prefer a passkey, which I
really must investigate.
And the site itself? It still doesn't supply information that I need, such as whether my
concession has been applied or not. Doubtless another ridiculously expensive upgrade that
does nothing useful.
Not quite 60 years ago I got my first
camera with electronics, an Asahi
Pentax“Spotmatic”. I
still have it. It looks exactly like the image in Wikipedia, down to the lens serial number:
As I commented at the time, the rearmost lens element protruded. This was what came to be
called the 8 element 50 mm f/1.4
Super-Takumar lens, and somehow the protruding element irritated me. When my father
also bought a Spotmatic 9 months later for the Asia
Trip, I discovered that the lens looked different, so I swapped it. That's the lens
in the image above, here from above:
This lens has seven elements, and the rear one (I think) contains Thorium, making it mildly radioactive.
I've just been watching a video about the various versions of this lens. I had thought that
the way to distinguish the models from outside was because it didn't have a click stop for
f/1.7, and also no marking on the aperture ring for f/2. But not all 7 element lenses have
that peculiarity. A more reliable way is where the infrared correction mark is on the
depth-of-field scale. The 7 element version has it between f/4 and f/5.6 on the scale, like
here, while the 8 element version has it to the right of the f/4 mark. It seems that the
change was made mainly for financial reasons, and the 8 element version was marginally
better. That's potentially borne out by the infrared correction. A pity I changed it.
Off to Cape Clear today for Pene Kirk to examine Larissa and give her an injection. She agrees that her condition is deteriorating, but that we
can keep her feeling well for some time to come, if necessary with more medication.
Also a discussion about what she saw a few days ago, for the first time in her career: a
male tortoiseshell cat. “There's no such thing”. And maybe that's correct: the tortoiseshell pattern comes
from X chromosome
inactivation, and for that the cat needs two X chromosomes—the hallmark of a
female. Pene agrees that it's probably an XXY genome.
I'm sure that I had a second package, and that I've eaten it already, but I don't seem to
have written anything about it. It doesn't help that greps for Pro bring up thousands
of false positives. OK, today's the day. The quantities are irritating, one cube for 500
ml. Normally I take between 300 and 340 ml. I reduced the water to 400 ml, which made the
broth rather too salty, but still not overly tasty. I think that's the last time.
Little technical pains
Topic: general, food and drink, technology, opinion
The batteries in Yvonne's salt grinder are weak and need
changing. Not a problem: we have dozens of NiMH batteries.
But getting the batteries out of the compartment proved to be a pain: they were in so tight
that I needed to use a screwdriver to lever them out. But that was nothing compared to
putting them back in: in one case it just didn't work.
Why? Clearly I have done this before. I bought the mills (a matched pair for pepper and
salt) in June, but it seems that I didn't mention it here; presumably I was too preoccupied
with the new dishwasher. They're somehow typical of their kind: one takes four batteries,
the other six. This one was the four battery version:
The compartments are so tight that I just couldn't get the last battery in. Why? I tried
other (non-rechargeable) batteries and couldn't get them in either, so it must be something
to do with the compartment. But it doens't look damaged. In the end I had to get a pipe
wrench to force it in. I wonder how long this device will last, and also what the
tolerances are for these things. They must be considerably less than 1 mm.
And while I was at it, there were a couple of minor things to do with the car: fill the
windscreen washer reservoir and check the tyre pressures. And they, too, were
a pain! First, the cover of the reservoir was stuck on, and I couldn't pull it off.
And to add injury to insult, the pneumatic bonnet lift support is old and worn out and kept
closing on me. Score one for the old prop rod method, which I simulated with a part for a
shelf. Bruno wanted to know all about
it, of course:
And a good thing I still had the pipe wrench handy.
And then there was the tyre pressure. The air pump is at the end of the garage, and I had
never got round to putting on the wheels. How hard can it be? A wheel, a rod and a nut.
So I tried. The nut is a locking nut, requiring some pressure to insert the threaded rod,
requiring some pressure, and it's in an inconvenient place. In the end I gave up:
And the tyre pressures? Yes, there's something wrong with the front right wheel, which
showed almost no pressure. I'll have to keep an eye on that one.
Yvonne wanted Jennys Lammtopf for dinner today. Problem: my recipe states 200 g of filling for me,
and 120 for Yvonne. And we only had 160 g. OK, give it to
her and I'll eat something else (curries that Yvonne doesn't like).
Problem: the numbers don't add up. The total was 80% of what I had for myself in the
recipe, but the pot was nearly overflowing:
The year is coming to an end, and with it the plans for my yearly annual newsletter. That should include a photo of
Yvonne and myself, along with as many animals as we can keep
still. Here last year's photo:
That was done with Hugin. But we
had decided to repeat the views every 10 years, so the coming photo should look something
like this, taken in December 2015:
OK, a clear-cut case for backgroung replacement. As planned last week, time to look at Affinity. There's quite an
understandable video on
the topic. First step is to load the image and choose the owl icon in the left-hand icon
column:
But my display wasn't like that. It seems to have been completely rearranged. The
left-hand column was only about half as wide, and all the incomprehensible icons had been
replaced by other incomprehensible icons:
And that's on a 1920×1080 display. On my normal 3840×2160 display the icons are only half
as wide again, about 3 mm. When I click on the third from the top, I get a message telling
me yes, this is the object selection tool:
But that's as far as I got. It told me to click on the object (myself in this case) and it
would be selected. But that didn't happen. Instead a tiny popup appeared elsewhere in the
image, so far from where I was working that I didn't notice it at first:
But then things seemed to start working by themselves. At least the cursor changed shape.
But I still couldn't select the part of the image. Off looking for answers: this is
Affinity 3, and all the videos refer to versions 2 or 1. I should have seen that from the
old video (9 months!) and the different logo. Written documentatation? That's an old,
worn-out magic word. After some time searching, I still couldn't find out how to do things.
In passing, I'm getting more and more fed up with the lack of scalability in “modern”
software, maybe because of my “high resolution” monitors that can display almost 10% of my
high resolution photos at a time. Why can't they scale to the size on the display rather
than to the number of pixels? My main displays have 167 pixels per inch, way more than the
75 or 100 that were once considered normal.
Gemma and Jess, surnames currently unknown, along to do the cleaning today. Yvonne was satisfied, and they cleaned away the dropped leaves from the
dying Hibiscus:
All software seems to do this nowadays, but this is the first that has made it so difficult.
OK, download the module, the only free one in the list, and try again. Success! Or what
passes for success nowadays: I was able to mark my profile and get a (very) rough mask,
which I could then refine with the “paint” tool—I think, though it left fringes that
wouldn't go away:
But the mask kept jumping around, selecting whatever the cursor was positioned on. Still, I
sort of got it to work. Next to add the background. But when I expanded the image, I
discovered that I had missed some of the misidentified area. OK, paint over it. But this
time I painted black:
And I couldn't find out how to change it. It was somewhere in the video, but I didn't want
to go through it all over again. And this was just a test run, so I left it like that.
Next in the video was to introduce the background. And I failed completely! The video is
not very clear how it does it, but it appears to involve first opening the image and then
moving the cursor to achieve anything, including adding to the composite image. And I made
the mistake of lifting the button while stretching it, so it ended up too small in the top
left corner:
And nothing I could do could change that. Finally found out how to delete it and tried
again. This time something else went wrong, and without any obvious action on my part it
was inverted!
ENOUGH! I tried Affinity because it's still being supported, but there is no documentation
that I can find, and the user interface is both bizarre and very different from previous
versions. There's more aversion than affinity there. Back to ACDSee Gemstone 12, which has the advantage of a number
of videos, all old, and many claiming that Gemstone 12 is in BETA. Interestingly, the
screen display looks surprisingly like the Affinity, though accessing files is less baroque.
But that was all that I had time for today.
Why am I having so much difficulty? I didn't have anything like this much trouble with
basic photo processing 15 years ago. Am I losing my marbles? It's a possibility, but I
don't think it's the case here.
My web sites are still frequently overloaded, with load averages sometimes over 200. And
the servers are issuing frequent ECONNRESETs. Once again I need to think about
what to do.
I already have code to delay and redirect long imagesize strings. Will that
help?
Yes! Though it's very dependent on the delay. At 30 s, the load averages dropped below 1.
With a bit of experimentation I discovered that 10 s was probably a good compromise, with
load averages round 2 to 10. And something that I didn't notice last time: after a
redirect, there was no second attempt. Is that a feature of the web server (automatically
serve a different page), or is it an indication that the crawlers already have the base page
(without imagesizes)?
Another breakfast experiment today. I really can't get my head around Japanese food.
Mainly to use up some ingredients (finally the cabbage is gone; it'll be a while
before I buy any more), I made this:
“China is a mess”. I wonder what it really means. But it seems that the content is more
than just miso.
The result? Edible. I had trouble with the udon, which didn't want to disentangle, and
the tauge (one of the reasons for the dish) was out of place. I really need to
understand Japanese food.
As planned, back to my investigation of ACDSee Gemstone 12 today. I didn't get very far. Once again I'm torn between videos that show you
what happens if everything goes to plan, and the lack of documentation to help when they
don't. Yes, it seems to do what I want. But I've already seen how badly Affinity recognizes shapes. Is there any reason to believe that Gemstone will do any
better, especially when they seem to share a common ancestry? And how do I tidy up
selections? I can't find anything to tell me.
I think we'll do the newletter photo the old way this year. No cats.
I've always had an aversion to codes of conduct. They should be unnecessary: people with
normal upbringing wouldn't offend against any reasonable code of conduct. The requirement
to commit to abiding by one (bound to be subtly different from the next) is tantamount to
being put on probation.
But now I hear that the Bendigo
Writers Festival almost failed this year because of a code of conduct: Bendigo Writers Festival
boycott. Fifty participants boycotted the festival because the code of conduct
included prohibition of criticism of Israel. Good for them.
Yesterday I decided to give up the attempts to perform background replacement with
the software available to me. So what did I do today? Attempt to perform background
replacement with the software available to me. Had I really used all the opportunities
available to me?
First discovery: the version, though claiming to be Gemstone (and not BETA), had significant
differences in the user interface. In particular, one of the items he chose is now (my
version of Gemstone 12) hidden one level further down beneath a different menu selection.
It's things like this, along with the far-too-fast selection of bewildering icons, that made
life difficult. The clip is about 4 minutes long, but it took me nearly an hour to get
through, sometimes frame by frame, and and the end it still didn't work.
Once again the similarity between Gemstone and Affinity Photo became apparent.
Selecting a foreground showed the selected part in red in the video, while my version had a
dotted outline—just what I saw with Affinity. I had to go back and confirm that I was
really using Gemstone. And while this time I was able to add a background, moving it was
exactly like showed in the Affinity tutorial.
Other irritations were setting mask colours. It turned out that I had missed a sub-second
part of the video where the presenter selected not one, but two bewildering icons.
Finally I had something like a mask. It wasn't perfect, though better than Affinity's. But
Yvonne's hand had been left out. Tried to fix it by painting
over it, but that didn't work, though I did find out how to change from black to
white: select the boxes at the bottom on the left
That's not an error on this page. The page contains only two lines (enlage to see anything at all), and the rest is blank. More serious,
though, is that I found nothing about background removal in the entire manual!
In the end, I found a way to save files—sometimes: c-s. But frequently it just
doesn't react. I still haven't found why. The best image I could get was:
Will I ever learn to get something useful? Arguably I should go back and look at Affinity
again; some of what I learnt here also applies there. But what a way to spend an afternoon!
My web site load continues to be very high, and it's clearly mainly crawlers. Does
Google Gemini know anything? “What is the
easiest way to identify a web crawler http request?”
Yes! The honest ones identify
themselves, but there are a whole lot of rogue bots that pose (via the User-Agent
specification) as an old browser. I can do without them:
$REQUESTER = $_SERVER ["HTTP_USER_AGENT"];
if (($REQUESTER == "Mozilla/5.0 (Windows NT 10.0; WOW64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/48.0.2564.116 Safari/537.36")
|| ($REQUESTER == "Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/109.0.0.0 Safari/537.36 Edg/109.0.1518.61") )
{
http_response_code (403);
exit (0);
}
That works. But to my surprise, instead of lowering the load average, it increased
it significantly. Why? I'm still trying to get my head around it.
Worse, I am still getting so many requests that the network stack can't keep up:
Dec 14 08:03:41 lax kernel: sonewconn: pcb 0xfffff8006a459b70: Listen queue overflow: 193 already in queue awaiting acceptance (704 occurrences)
Dec 14 08:04:43 lax kernel: sonewconn: pcb 0xfffff8006a459b70: Listen queue overflow: 193 already in queue awaiting acceptance (403 occurrences)
Dec 14 08:05:44 lax kernel: sonewconn: pcb 0xfffff8006a459b70: Listen queue overflow: 193 already in queue awaiting acceptance (719 occurrences)
That's up to 12 rejects every second! Clearly I must do something about it. Others,
like Daniel O'Connor, suggest firewalling the sites for a short period of time. That
requires significantly more effort, including potentially a new firewall and identification
software like Fail2ban. Let's
wait and see if the crawlers get tired.
I'm off to Werribee tomorrow,
1½ hours' drive each way. Yes, I can drive via Geelong, but I know that road well. How
about a more scenic route via the Brisbane Ranges?
Google, find me a route:
Oh. Different routes! Why did it do that? And none of the routes look “right” to me. The
northern routes go via Sebastopol, which is quite a diversion from the way through Meredith, the one that I wanted
to take.
But then I read the fine print (web browser only; we don't want to worry the pain phone
users): “Avoids road closure on Mt Mercer-Dereel Rd”. Oh. That's right where I would want
to go. So it looks like Geelong after all.
Off to Werribee today for a
periodontic examination. How do I get there? Normally I would have gone via Meredith, Victoria and
Anakie. Meredith is 34
km away, but the road closure on the Dereel—Mount
Mercer road added a full 21 km to the route. But the route via Bannockburn is boring, so I
chose it anyway, running into no fewer than four road works on the way, none of which had an
obvious purpose. Still, it was interesting, though I had forgotten to start the GPS Logger
that Android had stopped for me despite my express instructions, so all I got was this toy display from
Google Maps:
How did it compare with Google Maps' estimates? Same distance, two minutes longer. After
deducting the 6 minutes spent waiting for road works, it's actually faster. And presumably
it would have been round 1 hour if I had been able to go via the Mount Mercer Road.
One of the nice things about Google Maps are the images. But to my surprise, things looked
nothing like I expected. I should possibly look more carefully.
My appointment was with Leela Movva,
whom I have not seen for over 3 years. And though he no longer had my records (which remained in
Geelong), he recalled details
about me. A relatively short examination, in which he didn't quite match Mario's gum depth
of 9 mm, but did find multiple depths of 8 mm, still far too much. Dignosis: periodontitis. He has a solution:
laser therapy, which seems to be new in Australia, though practiced (and
hopefully perfected) in California for over 20 years. The costs are eye-watering, over $3000, from which I
discover that my health will refund the princely sum of $739.90. I agreed that I would talk
about it to Mario, but I suppose I will have to do it.
After the periodontics, I had planned to look round Werribee for an East Asian grocery, but I
couldn't be bothered. Off instead to the Fresh Land Asian Food in
Geelong, where I went in July. I had a fairly extensive shopping list, and I wasn't able to find
everything, but to my surprise I couldn't even find tauge. In many ways the Fruit Shack is better.
I did, however, find a number of noodles I hadn't seen before:
More and more, the inability to read Chinese is becoming a problem. Today I took a couple
of photos of jars that interested me, using my Olympus E-PM2:
The first are something like pickled beans—I bought the right hand two. But what's the one
on the left? To be examined when I can get a legible translator. And I went to a lot of
trouble to confirm that it was in focus. So when I took the second photo, I didn't bother.
The sides are in focus. How did I mis that? Off-centre? With a real viewfinder, I would
have noticed.
There's no question, the grid power reliability has increased considerably in the last few
years. Apart from far too many “planned outages”, there has only been one signifcant outage
in the last 15 months.
Until today. And once again our PV inverter showed itself from its worst side by powering down for about 15 seconds.
Time to see how my new UPS in the office
works.
Wonderful! No interruption. Oh, well distress, my Microsoft computer, was running
at the time, and it's not connected to the UPS, so it rebooted. But all the other machines
were on UPS, and they carried on running.
But then Yvonne came to me and told me that her mouse wasn't
working. Neither was lagune, her machine. Why? I thought it was still running
after the failure, but the last message in /var/log/messages was from 40 minutes
before the failure. That in itself suggests that it's an unreliable source. But why
should lagune have failed? It's behind a UPS like all the others.
Worse: after rebooting she couldn't recover firefox. For some reason she had a different profile, and it was relatively
easy to recover.
But she couldn't access her mail! And here again she had a different configuration: it was
looking for /usr/mail/yvonne instead of /var/mail/yvonne. Why? There was no
obvious change in the configuration. And while searching for the reason, I discovered that
it happened only under X. Why? Put it into the
“later” basket and addressed other issues of the day. A brief look
at ~yvonne/.xinitrc showed some issues, though: for lagune it had
xpat is a graphic card game, and it came up. But xearth didn't, and clearly
the instance for lagoon:0.1 is nonsense. But I was able to start xearth from
a shell. When I have time I'll need to look through the log files in detail.
Oh. /home/local/bin/xearth. That doesn't exist (any more). Another edit.
I still need to look at those log files.
Then later I wanted to watch TV. Problem: teevee still has strange display problems,
so I can't return to the console. Instead I start X over the network. And for some reason Xorg:0 stopped, though Xorg:1
continued running, but accepted no input. So I had to restart it. But why? The connection
goes through no fewer than 3 switches that briefly lost power, but a 15 second outage
shouldn't have caused it to stop. Again something for the “later” basket.
And then, just before going to bed, I got the first of my nightly backup mails. But instead
of the normal 100 kB or so, it was 110 MB! Looking, I saw:
DUMP: Dumping /dev/nvd0p2 (/) to standard output
/home/local/bin/dodump: cannot create /dump/hydra-FreeBSD/2/root.bz2: Input/output error
DUMP: mapping (Pass I) [regular files]
...
DUMP: Broken pipe
DUMP: The ENTIRE dump is aborted.
mkdir: /dump/hydra-FreeBSD/2: Input/output error
/home/local/bin/dodump: cannot create /dump/hydra-FreeBSD/2/tars/dumpdate: Input/output error
cat: /dump/hydra-FreeBSD/1/tars/dumpdate: Input/output error
(thousands of repeats)
What's wrong there? /dump is on eureka. so it's all over NFS. Into the office to
discover that, although eureka itself survived the power outage, the external USB
disk power supplies lost power. Not only did they power cycle, once again they came up with
different device numbers. Tried with moderate success to force unmount and remount the
first disk, but the second umount hung with a strange situation. It wouldn't respond
to kill -9, of course, but top showed me:
PID USERNAME THR PRI NICE SIZE RES STATE C TIME WCPU COMMAND
63712 root 1 20 0 12384K 1972K CPU3 3 0:00 100.00% umount
100% CPU time and no time used? That makes no sense. I had expected it to be waiting on
something. OK, what does ktrace say? Nothing. A completely empty dump file. But a
little later I saw the CPU time suddenly increase to 3:18. My guess is that it was spinning
somewhere deep in the kernel. And gradually other processes ended up waiting on whatever it
was holding, even on hydra.
What to do? For once, the Microsoft solution: reboot. Even that didn't go completely
smoothly: logging in as root on /dev/ttyv2 worked normally, but logging in
as myself on /dev/ttyv3 hung, though it recovered in the course of the night.
Logging in via ssh worked, however.
Leftover duck from Saturday today, just what I need for a breakfast of Dereel duck noodles.
Oh. That was 3 years ago, and since then, with the exception of the duck, it has basically
become my generic breakfast noodle dish. OK, this time just use duck and the “marinade” in
which it came, along with choi sam, dòufu and tauge:
Bondi Beach is well known in
Australia as one of Sydney's best
beaches. But now it's known all over the world for the massacre that happened on
Sunday: two terrorists, father and son, attacked a Hanukkah celebration and killed 9, no 11,
no 15 people. Those are the kind of thing that almost only happens in the USA.
Who is to blame? I pointed the finger a year ago: Benjamin
Netanyahu. But instead he blames Anthony Albanese. Bibi, Jews are
dying because of you (not to mention the inhabitants of Gaza). When will this ever stop?
In passing, it seems that the murderers were cowardly. That's a term that the press
is using more and more, like “graphic” to describe not graphics, but distasteful images. I
would have thought that any coward would never even consider mass murder. Where did this
usage come from?
Another Zoom meeting
with Aaron Harivel of OM
Digital Solutions today, on the topic of beginner birding theory. That seems out of
place, since I've been birding with him in person a couple of times, but past experience
shows that he might have something interesting.
Things didn't start off well. For some reason, my video camera (looking at me
on eureso, the laptop) didn't work correctly. I wasn't the only one: a number of
other people showed up on the list without video, and Aaron had an unexpected disconnect
right at the beginning. And I couldn't find a settings link on the laptop (standalone)
version, though the one on hydra had settings but no camera.
Apart from that, he had a number of suggestions, some of which he had told me earlier, but
which I hadn't written down. Today I took lots of screen shots and made some notes:
Use the small square focus on the E-M1.
He mentioned a colleague who used aperture priority on his birding. I still need to get
my head around the reasoning.
Apply the kind of subject detection to a dedicated button. Again, the screen shots
might help. He uses the ± button on top of the camera.
Subject detection disables tracking.
Use the focus limiter.
Mechanical shutter doesn't produce banding
For ProCap, remove the capture counter.
From one of the participants: "Also, suggest looking for Ramsar sites and seeing if
there’s any near you". It seems that Australia has 67 Ramsar sites, covering 83,000 km²
(equivalent to a squre 288 km on a side), and one of the interesting ones for me are the
Western District
Lakes, which I must investigate
Mobile connect, the link between camera and mobile phone that I thought completely
useless: it could good for birding.
He discussed teleconverters and why they only work on certain lenses: they have a nose
that protrudes into the lens, so a lens with an element in the extreme back won't work.
Took a look at my Leica DG Vario-Elmar 100-400 mm f/4.0-6.3 and didn't see anything until nearly 5
cm into the lens. Will it work? No, says Aaron, but Panasonic make their own
teleconverters. It wasn't until later that I discovered that I had not seen the
reflection of the rear element: it's right at the back of the mount. And the Panasonic converter (TC20) costs an arm and a leg.
For that price I could buy a second-hand long lens.
Do you have a comment about something I have written? This is a diary, not a
“blog”, and there is
deliberately no provision for directly adding comments. It's also not a vehicle
for third-party content. But I welcome feedback and try to
reply to all messages I receive. See the diary overview for more details. If you do
send me a message relating to something I have written, please indicate
whether you'd prefer me not to mention your name. Otherwise
I'll assume that it's OK to do so.
