Twitter: @ewenmcneill -- November 2020
Sun Nov 01 20:40:59 +0000 2020 (#)
RT @bunniestudios: Today we're launching the campaign to fund Precursor, a mobile open source electronics platform for developing secure ap…
Mon Nov 02 20:40:11 +0000 2020 (#)
RT @linuxconfau: Yikes! It’s the last week to submit your session and miniconf ideas for http://linux.conf.au 2021. If you have ideas or…
Tue Nov 03 00:24:12 +0000 2020 (#)
So far my impression of “serverless” computing is that my 1980s 8-bit computer had a much faster edit, compile, test, debug cycle, using floppy disks. And it could report more than one error at a time. Even some of them at compile time instead of run time.
Tue Nov 03 00:25:46 +0000 2020 (#)
Replying to @ewenmcneill
Not really sure what the point of “validate-template” is if it doesn’t even report that required parameters are missing. (And deploy seemingly can only report one error per deploy attempt 😢)
Tue Nov 03 03:53:21 +0000 2020 (#)
Replying to @unixbigot
Yes, this is where I’m up to now. Debugging the gap between the slow local version that works (Docker emulation of “serverless”) and what actually got deployed to the cloud to run. It appears the difference between two deployment methods matters here...
Tue Nov 03 03:55:08 +0000 2020 (#)
Replying to @unixbigot
But I’m at a point where it took less time to write a (partial) 8-bit CPU emulation in Python (as an extended joke) than it’s taken just to get this deployed and (not) running in the cloud.
Which feels wrong 🤔
Tue Nov 03 04:16:12 +0000 2020 (#)
Replying to @hroethgar
IDK, programs for 1980s 8-bit computers tended to be very locked in to the target platform too...
... even CP/M which provided a little bit of abstraction soon leaked implementation details like terminal control sequences out to the application developer 😃
Tue Nov 03 05:30:57 +0000 2020 (#)
RT @devonzuegel: Why do we call them "open source developers" when we could call them "open sourcerers"???
Tue Nov 03 06:55:45 +0000 2020 (#)
Replying to @aurynn
TBH I’d say an airplane is a place if you are inside it (you’re there, it’s a place). And an object if you are outside it, viewing it.
I’m not sure that makes the edge cases any easier though 😃
(An ocean is a place, by the same definition: if you’re in/on it, it’s a place...)
Wed Nov 04 07:06:47 +0000 2020 (#)
It turns out that the original #Covid19NZ order requiring QR codes to be displayed has been revoked. And the order that revoked it has been revoked. I think the current version is “COVID-19 Public Health Response (Alert Level Requirements) Order 2020”.
http://www.legislation.govt.nz/regulation/public/2020/0231/latest/whole.html#LMS400549 https://twitter.com/ewenmcneill/status/1293339789056290816
Wed Nov 04 07:09:39 +0000 2020 (#)
Replying to @ewenmcneill
The “latest” version I found via the #NZCovid19 “Legislation and key documents” page, and it doesn’t seem to have been revoked yet.
But TBH such a key requirement really ought to have been given its own permanent home, not just in random revocable orders. https://covid19.govt.nz/updates-and-resources/legislation-and-key-documents/
Wed Nov 04 07:18:10 +0000 2020 (#)
Replying to @ewenmcneill
Of note the latest version applies to a “workplace”. Which arguably even more strongly includes home offices than the earlier version. (And on some readings maybe includes everywhere a trades person visits to do work...)
http://www.legislation.govt.nz/regulation/public/2020/0231/latest/whole.html#LMS400544
Wed Nov 04 07:25:40 +0000 2020 (#)
Replying to @ewenmcneill
Note that most of the “COVID-19 Public Health Response (Alert Level Requirements) Order 2020 (LI 2020/231)” (linked into tweet above) has also been revoked. But AFAICT not the requirements around QR codes.
Which is a really confusing way to handle things 😢
Wed Nov 04 07:28:32 +0000 2020 (#)
Replying to @ewenmcneill
It’s really help if the #NZCovid19 page saying “this is a legal requirement” actually linked to the current legal requirement, rather than just asserting it was true.
As it’s unreasonably difficult to find the current legal requirement definitions :-(
https://covid19.govt.nz/business-work-and-money/business/get-your-qr-code-poster/#who-needs-to-display-a-qr-code
Wed Nov 04 18:57:28 +0000 2020 (#)
RT @herrprofdr: For students in my now-online retrogame class next term, I built an in-browser 6502 debugger atop the @internetarchive's Em…
Thu Nov 05 07:23:23 +0000 2020 (#)
This lovely arrow anti-knight sudoku turns into a beautiful colourful grid in about 45 minutes. Lots of fun to watch the “sudoku algebra” expressed in colours, which all get resolved in the last few minutes!
💜💚💙❤️🌈❤️💙💚💜 https://youtu.be/Jk2ko0-eP7c
Thu Nov 05 07:56:52 +0000 2020 (#)
RT @merxplat: Wellington look at this
Thu Nov 05 22:36:53 +0000 2020 (#)
Replying to @ewenmcneill
Apparently one of the main reasons that local testing is so slow is that they don’t cache local layers, only downloaded ones. So it constantly rebuilds the local files, taking tens of seconds each time, because it’s hard to tell if a local file changed 🤦🏻♂️ https://github.com/aws/aws-sam-cli/issues/1331#issuecomment-520047758
Fri Nov 06 20:14:36 +0000 2020 (#)
Thread looking inside a 400G AOC (ie 16-core fibre linked SFP pair). Looks like there’s a MPO connector on the outside of each SFP, for easier fibre cable routing. https://twitter.com/KWF/status/1324402672833757184
Sat Nov 07 03:57:16 +0000 2020 (#)
ComfyCon AU Easter was great, so I’m expecting the Second Edition to be fun too.
Online, based somewhere East Coast, Australia, so I believe running on UTC+11; with a planned noon-21:00ish UTC+11 schedule, it’s very “sleep in” hacker con hours in NZ (14:00ish-23:00ish UTC+13). https://twitter.com/ComfyConAU/status/1324881661067055104
Sat Nov 07 03:59:53 +0000 2020 (#)
RT @ktemkin: everyone: stop acting like women’s bodies are somehow fundamentally shameful / inherently inappropriate
it’s not on us to hid…
Sat Nov 07 04:15:52 +0000 2020 (#)
RT @ktemkin: but watching other women being censored/censured/forced to-self-censor is more than I can silently bear
so to hell with the i…
Sun Nov 08 07:39:48 +0000 2020 (#)
Replying to @TProphet
?
$15-$18 for a cocktail is pretty standard here (NZ) too. At least here a non-trivial percentage is the tax on alcohol, which gets pretty high on high alcohol spirits. I’d guess it’s similar in most AU states.
“Duty free” on spirits made a big difference (up to 30-40%) 😃
Sun Nov 08 08:00:06 +0000 2020 (#)
Replying to @TProphet
If it helps, those AU prices (and the NZ prices) include sales tax too (GST). And cover paying the employees properly too (ie no “compulsory” service charge or tip required on top of that).
My memory is “US prices plus sales tax plus 20% tip” are closer than sticker prices :-)
Sun Nov 08 20:07:27 +0000 2020 (#)
Replying to @thatcks
My UPS does that too. I think it’s the buck/boost circuit of the Line Interactive UPS kicking in, when it happens without the UPS alarm sounding (and hence no alerts).
It used to do it a lot just before the supply line fuse got replaced, now it’s pretty infrequent.
Mon Nov 09 05:00:36 +0000 2020 (#)
Replying to @ktemkin
For this reason, I’d use “wife” as a shorthand from now on. With clarification of <<“de facto”/“common law” wife, engaged to become “de jure”/“legally married” wife>> if the context requires it.
But “wife” is wonderfully expressive here, and I suggest adopting it immediately :-)
Mon Nov 09 05:03:55 +0000 2020 (#)
Replying to @ExcitedLeigh
Is “leave it on the footpath with a sign saying ‘free to a good home’” a thing in Your Town? Because that seems to be a common way to handle “I just need this to be gone” in a bunch of places.... 🤔
Mon Nov 09 07:13:58 +0000 2020 (#)
Replying to @LapTop006
Photoshop, select tight bounding box on text, copy selected boxes to new layer, magic wand select text on that layer (shift-select with 3x3, tolerance 5), copy selected text to two new layers, multiply second layer against first at about 80% opacity.
Mon Nov 09 07:20:29 +0000 2020 (#)
Replying to @LapTop006
It may also help to stretch contrast in the text layer before duplication (and crush the light greys to white) before multiplying. Looks like that’d need 70% opacity on the multiply group too, as it’s quite intense. Add a layer group mask to tidy up weirdnesses.
Mon Nov 09 07:21:16 +0000 2020 (#)
Replying to @LapTop006
Nice street art capture BTW :-)
Mon Nov 09 08:48:14 +0000 2020 (#)
Replying to @LapTop006
Also, I’m reminded of “I break rule 6” (hugs and back rubs) from the good old days of Usenet. Which I see has been used as a signature through until at least 2018!
I even remember the 1991 post explaining it. Nearly 30 years ago.
Hugging is subversive! https://web.archive.org/web/20180830210834/rule6.info/timecon.html
Mon Nov 09 09:09:10 +0000 2020 (#)
Replying to @LapTop006
There’s definitely a whole bunch of random signs this year that would be so out of context In the last decade...
I don’t usually do street photography, but did try to capture some of the ephemeral signs, as they rapidly vanish again.
(Eg some from NZ level 3 lockdown)
Tue Nov 10 05:22:03 +0000 2020 (#)
RT @sigmas: OMG, that #Seattle #sunrise this morning!!! Sky was on fire!!!
Tue Nov 10 09:11:14 +0000 2020 (#)
Replying to @ktemkin
Pretty much all code is wrong at fractal levels of detail, all the way up and down. I try not to think about this too much... 😃
(Sometimes it’s wrong in the right ways: the world seems to run on bug for bug compatibility!)
Wed Nov 11 07:24:25 +0000 2020 (#)
RT @OlofKindgren: @sylefeb @BrunoLevy01 Thank you! Happy to hear. And if you're ever in possession of 7 spare minutes, there's one about SE…
Thu Nov 12 07:52:28 +0000 2020 (#)
TIL the #NZCovid19 app is still auto logging me out (despite my previous use being only a couple of days ago).
But it finally prompts for iOS password store save, and when granted will save the password for future use with the normal iOS authentication. So that’s progress...👍
Thu Nov 12 07:55:15 +0000 2020 (#)
Replying to @ewenmcneill
The fact that it’ll also scan QR code’s without being logged in (a change a couple of months back) also helps; I did at least get to use the app this trip (before getting back to my main password store).
Thu Nov 12 07:59:15 +0000 2020 (#)
Replying to @ewenmcneill
But since none of the scans sync’d onto the app on my new phone when I logged in there (to save the password there too), I still wonder both what the point is of having an account at all and the point of auto logging people out. Especially if the scans just stay on one phone.🤷♂️
Thu Nov 12 08:17:30 +0000 2020 (#)
Replying to @TProphet
Microsoft is probably tied to Intel for as long as the PC OEMs are, which I’d guess is at least 5-10 years under existing contracts.
None of those (Intel, OEMs) have the vertical integration (silicon to OS) that Apple has to fix integration problems at the easiest level.
Thu Nov 12 08:21:29 +0000 2020 (#)
Replying to @TProphet
But I could see one of the Clouds (Google first maybe?) going to ARM or RISC V on their own silicon designs.
Which could make a huge difference to Intel as they’re all very big compute hardware purchasers...🤔
(Seems likely Apple’s Cloud would go ARM over next few years too.)
Thu Nov 12 08:30:01 +0000 2020 (#)
Replying to @ktemkin
Kingdom. With all it contains.
Possibly only a small Kingdom for $10M, but maybe you can use that as a down payment for a leveraged buyout, and sell off the bits you don’t want to cover the rest of the costs... 🤔
(Also $10M seems to be the going rate for a Knighthood.)
Thu Nov 12 22:38:08 +0000 2020 (#)
Error: success 🤦🏻♂️
Dear Microsoft, if you can’t find a session then the user is successfully signed out. Just saying. (Eg, the signed out in another window or the session auto expired.)
Fri Nov 13 05:36:50 +0000 2020 (#)
The beautifully symmetrical visual design of this Sudoku puzzle is only the beginning of the symmetry revealed in solving the puzzle!
Simon (of @crypticcracking) makes it even more beautiful with clever use of colouring during the solve :-)
Great solve! https://youtu.be/VfTFtnCBtOQ
Fri Nov 13 06:43:44 +0000 2020 (#)
RT @kenshirriff: Here are the two dies at the same scale. The M1 is over twice as large physically as the ARM1. It has 16 billion transisto…
Fri Nov 13 09:38:05 +0000 2020 (#)
Replying to @ExcitedLeigh
The ratio of terms and conditions size to window in which to read them seems to be asymptotically approaching infinity...😂
(Since the only term that matters is “we can change these terms and conditions at any time”, I guess reducing the window to “I agree” has little effect 😢)
Fri Nov 13 19:02:15 +0000 2020 (#)
Replying to @techpractical and @pjf
Crowdfunding now :-)
(Technically it’s been crowdfunded for years. But The Rick Astley Hotline could possibly do with a funding boost this week 😃)
https://www.patreon.com/_pjf
Sun Nov 15 06:02:11 +0000 2020 (#)
RT @OlofKindgren: I have enjoyed every single one of the Dial-Up episodes and highly recommend watching them if you want to learn more abou…
Sun Nov 15 06:31:39 +0000 2020 (#)
RT @kenshirriff: A DRAM memory chip stores bits in tiny capacitors. I reverse-engineered the popular MK4116 chip (1976). It holds 16,384 bi…
Sun Nov 15 06:40:19 +0000 2020 (#)
TIL that DRAM row/column layout got influenced by the Z80s internal DRAM fresh feature, which apparently wasn’t big enough to cover 256 rows 😮
(“My” Z80 based first computer — Amstrad CPC6128 — let the video scan out, which covered 16kB per frame, trigger the DRAM refresh.) https://twitter.com/NotThatGreg/status/1327675442367750144
Sun Nov 15 20:52:27 +0000 2020 (#)
Replying to @NotThatGreg
That’s clever (CPU driving the DRAM cycle), and quite the kludge (CPU spends most of its time running NOPs)!
The CPC had a 6845 CRTC driving the video cycle and a Gate Array mediating access to DRAM. IIRC they just stalled the Z80 during CRTC RAM access, to share access. 1/2
Sun Nov 15 20:54:56 +0000 2020 (#)
Replying to @NotThatGreg
Since the Z80 doesn’t access DRAM on every clock cycle (roughly 2 clocks out of 4, depending on instructions) this mostly meant the Z80 ran unimpeded with 6845 and Z80 DRAM access alternating. But the 6845 always got right of way 😃 2/2
Mon Nov 16 08:19:21 +0000 2020 (#)
Replying to @minxdragon
So far, of the people I’ve seen posting this picture in the last couple of days, it seems like the most common gender is online...
(Possibly selection effect 😂)
Mon Nov 16 09:13:43 +0000 2020 (#)
It turns out that 1.6TB of USB3 attached encrypted disk copied from within the encryption to another USB3 attached disk takes about 31 hours to copy, at 14MB/s and 55-60% CPU usage (apparently a single core). I guess it was CPU throttled.
Mon Nov 16 09:16:02 +0000 2020 (#)
Replying to @ewenmcneill
This is the encrypted Time Machine drive corrupted by a macOS patching cycle a few weeks back. After much procrastinating I decided to make a copy of it before I erased the partition and started Time Machine again in that drive (meanwhile my other TM backup target still worked).
Tue Nov 17 06:20:36 +0000 2020 (#)
Imagine a 4D Sudoku puzzle, boggle a little at the concept, then watch Simon (of @crypticcracking) solve the puzzle!
It’s a fun 35 minute solve if you like your puzzles to be brain teasers! https://youtu.be/Nxj_VA8gzaU
Tue Nov 17 07:21:28 +0000 2020 (#)
Replying to @ewenmcneill
FTR the dd copy to an unencrypted partition does mount read only (“cannot be repaired” like the original encrypted one). Which I expected. But the mount failure comes much quicker than I expected (I guess fsck of an encrypted FS is CPU bound on decryption too 🤔).
Tue Nov 17 07:23:54 +0000 2020 (#)
Replying to @ewenmcneill
So there’s at least a chance it might work for read only Time Machine restores for older files. Which is what I was hoping by cloning it before erasing it.
(The original, now corrupted, encrypted drive is going to be dropped from TM, erased, and then have a new TM setup on it.)
Tue Nov 17 21:29:02 +0000 2020 (#)
Replying to @ktemkin
“I pray that this shall come to pass” 😂
Tue Nov 17 23:33:10 +0000 2020 (#)
RT @chrisjrn: Things I've learned about Code of Conduct enforcement from 10 years of community development:
- Incident reports happen in…
Wed Nov 18 08:35:35 +0000 2020 (#)
Why not solve an entire sudoku with just colouring and partial orderings, then fill in the numbers at the very end, once you have a total ordering? Why not indeed! Fun solve (but note it takes an hour to complete!)
💚💙❤️💜🌈💜❤️💙💚 https://youtu.be/qFLD2TFijJQ
Thu Nov 19 07:37:41 +0000 2020 (#)
Replying to @ExcitedLeigh
FWIW the second NZ lockdown (level 3/2, then 2), a month or two back, was roughly 10 days plus another 10 days. Triggered by a similar number of cases to SA now.
With good compliance and careful source tracing/contact tracing, 6+8 days may work. But it’s definitely a minimum.
Thu Nov 19 18:58:04 +0000 2020 (#)
Replying to @thatcks
Is it maybe being caused by a switch from GPU rendering (composition?) to software rendering? Firefox is particularly bad for excessive graphics operations and probably relies on GPU rendering to perform okay. Most GPUs can only render to certain size viewport.
Thu Nov 19 20:45:49 +0000 2020 (#)
Replying to @thatcks
Definitely feels like a change in rendering path between the two cases. Hopefully you can figure out the cause, and ideally find a work around. (My memory of fvwm from years ago is it used “larger than physical screen” coordinates, and some GPUs have weird limits in that case.)
Fri Nov 20 02:48:03 +0000 2020 (#)
Replying to @ExcitedLeigh
This seems to be a problem in NZ too: contract tracing details end up being partially public, and then analyzed in detail by journalists/the public for “good”/“bad” behavior. Which seems very to cause details to be “forgotten” by others talking to contact tracers in future.
Fri Nov 20 02:50:33 +0000 2020 (#)
Replying to @ExcitedLeigh
Doctor / Patient (and Lawyer / Client) confidentiality exist not in isolation, but because over time societies found it more valuable to have them, and have people speak more freely so the problems could be addressed, than to have details hidden.
This is another case like that.
Fri Nov 20 03:38:48 +0000 2020 (#)
Replying to @ExcitedLeigh
TBH “asynchronous written text” counts as a “more expressive medium” to me (compared with real time text). If nothing else the delay and time to clarify wording can both help process the intent better.
Maybe voicemail/video mail could have similar effect? 🤔
Fri Nov 20 06:30:31 +0000 2020 (#)
Replying to @nickzoic
Video mail, you mean?
I think there’s a fair difference in production values between a conference talk for the public and a short “just wanted to send you a smile and explain” recording for one person you know well....
Sat Nov 21 05:59:32 +0000 2020 (#)
RT @greg_jenner: When I was 8 I was sure this was the greatest ever song. Then I learned more about music and realised it’s silly to say th…
Sun Nov 22 06:38:47 +0000 2020 (#)
Replying to @attacus_au
Deep Purple.
And I agree with @.ExcitedLeigh about using a very traditional font.
Mon Nov 23 07:08:07 +0000 2020 (#)
Replying to @TProphet
“Rude” is always relative to group norms/communication protocol standards (in network protocols we call it something else, but it’s functionally the same thing).
So I think at least a tacit group agreement to “skip tree idle time retrain” helps.
Mon Nov 23 07:10:46 +0000 2020 (#)
Replying to @TProphet
In both in person and remote voice Informal meetings I’ll tend to either continue talking or stop talking depending on my perception of my need to say the thing (eg avoid 10 minute detour when I have answer) or other person’s need to talk same, plus are they “under heard”.
Mon Nov 23 07:13:30 +0000 2020 (#)
Replying to @TProphet
In remote voice meetings were there’s a clear meeting lead I’ll generally say only that person’s (or relevant other person’s) name and then wait for an idle channel / to be called on.
Definitely helps to be extra aware who isn’t being heard in all these cases though.
Mon Nov 23 07:21:43 +0000 2020 (#)
Replying to @TProphet
Also, finishing your bit with “sorry NAME you were saying” to explicitly hand the channel to the person who was trying to talk when you were (rather than letting someone else steal it in replying to you) can be a useful way to ensure they get a clear chance to speak.
Mon Nov 23 07:39:18 +0000 2020 (#)
Replying to @tveastman
I’m also told (via someone who was considering travel to their overseas family) that most of the NZ incoming quarantine slots through to about the end of January are already booked out.
Which suggests there need to be more slots over the summer/holiday peak.
Mon Nov 23 07:42:51 +0000 2020 (#)
Replying to @puck_ and @tveastman
TBH less than $1/taxpayer/day sounds like money well spent to avoid getting more new community spread during a pandemic. Just saying.
https://twitter.com/radionz/status/1330575689968336896
Mon Nov 23 23:14:07 +0000 2020 (#)
RT @slyall: We are happy to announce the Sysadmin Miniconf has been accepted to http://Linux.conf.au 2021 . This year's conference is al…
Tue Nov 24 20:59:29 +0000 2020 (#)
If you’ve ever wanted to attend http://linux.conf.au, but the costs (travel, registration, accommodation, etc) have put you off this is the year for you: affordable tickets, and the conference comes to your existing accommodation :-)
Sat 23-Mon 25 January, UTC+11 (AEDT). https://twitter.com/linuxconfau/status/1331033726159970304
Tue Nov 24 22:12:50 +0000 2020 (#)
Call for presentations at the #linuxconfau #LCA2021 Systems Administration Miniconf is now open (closes Fri 2020-12-18 AoE).
Miniconf held online only, Saturday 2021-01-23 daytime UTC+11 (AEDT).
Come tell us your 2020 admin war stories! https://sysadmin.miniconf.org/cfp21.html
Wed Nov 25 02:55:33 +0000 2020 (#)
Well that was rather more “fun” than I wanted today.
RHEL 8.2 ships with krb5 1.17; RHEL 8.3 ships with krb5 1.18.
krb5 1.18 drops DES support and bumps the KDC database major version. So krb5kdc will not start.
(Rolled back to 8.2 snapshot.) https://web.mit.edu/kerberos/krb5-1.18/
Wed Nov 25 02:58:48 +0000 2020 (#)
Replying to @ewenmcneill
Oh and there’s at least one CVE in the krb5 1.17 that came with RHEL 8.2. So that’s a reason to upgrade too (fixed 1.17.2 and 1.18.3 apparently).
(I also found CentOS and Fedora reports; but not finding clear recommended upgrade path.) https://access.redhat.com/security/cve/cve-2020-28196
Wed Nov 25 03:01:18 +0000 2020 (#)
Replying to @hroethgar
Yeah :-/
Both from MIT Kerberos and from RedHat.
It’d be nice if the MIT Kerberos release notes maybe mentioned they bumped the database major version. And perhaps explained how to migrate it. Instead they’re just “yeah, we dropped DES” 😔
Wed Nov 25 03:03:31 +0000 2020 (#)
Replying to @hroethgar
Fortunately I had a VM snapshot from just before the upgrade attempt, as I didn’t trust the upgrade size. And it’s in a cluster anyway. So other than people being pinged by monitoring it was as easy as possible to roll back.
(Next upgrade attempt will Need A Plan.)
Wed Nov 25 03:55:48 +0000 2020 (#)
Replying to @ewenmcneill
Interesting. It looks like the KDC database version change happened about a year ago in the development branch. Not directly related to dropping DES (which has been a multi year transition).
Still... not the best way to handle KDC DB version bump. https://github.com/krb5/krb5/commit/7196c03f18f14695abeb5ae4923004469b172f0f
Wed Nov 25 08:11:11 +0000 2020 (#)
RT @timonsku: Really awesome interview by @glowascii with @bunniestudios about the Precursor https://www.youtube.com/watch?v=4VrPQvbUzhE
Wed Nov 25 09:30:34 +0000 2020 (#)
Great thread ⬇️
Successfully optimising any system requires a deep understanding, of the system as a whole, of the tradeoffs involved in the optomisation choices, and how the system feedback is interconnected. (Plus realising the act of measuring changes what you are measuring.) https://twitter.com/nickstenning/status/1331508676516077568
Wed Nov 25 20:23:23 +0000 2020 (#)
This Precursor update on how their PCBs are made is actually a fascinating tour of PCB manufacturing for modern high density devices (blind and buried vias, custom shapes, etc), with pictures of each of the stages.
https://www.crowdsupply.com/sutajio-kosagi/precursor/updates/how-we-make-our-custom-pcbs
Thu Nov 26 02:05:51 +0000 2020 (#)
Replying to @pjf
I feel like there’s a generation gap between those who used voice phone calls (and no voicemail) as their IM, and those who used asynchronous messaging (from email onwards) as their IM.
I think the latter are way more likely to put the actual request in the first message.
Thu Nov 26 03:49:45 +0000 2020 (#)
Replying to @eevee
I tend to either ignore those (if they don’t cover the part of the page I need), or use “reader mode” (if that let’s me see what I need), or use the Inspector in Developer mode to delete the div containing the overlay.
Kobayashi Maru 😂
(And Cookie AutoDelete too.)
Thu Nov 26 03:51:49 +0000 2020 (#)
Replying to @eevee
Maybe you can “set cookies”. But don’t expect to get them back again later (currently “more than 5 minutes after I close the tab” with very few exceptions).
(Plus of course the usual browser third party cookie blockers turned on too.)
Fri Nov 27 01:05:20 +0000 2020 (#)
RT @linuxconfau: Would you like to talk at #lca2021? We have four exciting Miniconfs that are now accepting proposals. Get your sessions in…
Fri Nov 27 08:52:08 +0000 2020 (#)
RT @ricphillips: Music nerds will love this - five short pieces in the under-appreciated locrian mode https://youtu.be/el1ZhkN85Mc
Fri Nov 27 22:22:10 +0000 2020 (#)
RT @GyledC: @ComfyConAU SE is happening this weekend #Australia timezone (UTC+11)! https://au.comfycon.rocks/2020SE/schedule https://discord.gg/78sNVF36XG Watch at…
Sat Nov 28 07:31:51 +0000 2020 (#)
Speaking about Shadow IT now at #ComfyConAU is @attacus_au in her persona as an omnipotent space being.
(No I didn’t colour or edit those pictures, just crop them; but I think Zoom had a hand in the beautiful space presence 😃)
Sat Nov 28 07:43:24 +0000 2020 (#)
Replying to @attacus_au
As always this @attacus_au talk rewards a second viewing (she also gave a version at PyConlineAU 2020). But the #ComfyConAU version has Zoom Enhance visuals (phasing in and out of this world), as an added bonus 😍 🌕🌖🌗🌘🌒🌓🌔
Sat Nov 28 08:07:03 +0000 2020 (#)
Replying to @attacus_au
A reaction meme to @attacus_au’s talk from the #ComfyConAU discord channel.
HatTip to syngularityO on Discord
Sun Nov 29 00:49:05 +0000 2020 (#)
Replying to @ExcitedLeigh
Or crop the photo FTW, instead of leading with the rejection, maybe Google?
It’s a nice photo, just with a bit of clutter around the edges. (Spontaneous photos often need a wee bit of cropping to focus on the subject of the photo.)
Sun Nov 29 00:53:31 +0000 2020 (#)
Replying to @ExcitedLeigh
Also the tweet directly below yours on my timeline was @.pjf pointing out developers don’t have to use ML to make the service worse. Which seemed apt.
Sun Nov 29 01:07:04 +0000 2020 (#)
RT @pjf: Just because you can use machine learning to make your product worse doesn't mean you have to.
Sun Nov 29 05:20:26 +0000 2020 (#)
TIL that there is a Markdown/Python/jinja2 based report generator, which outputs to PDF (via LaTeX) and HTML. Aimed at InfoSec reporting (with lots of common templated sections), but I expect it can be repurposed for other report types too.
#ComfyConAU https://gitlab.com/volkis/report-ranger
Sun Nov 29 08:15:21 +0000 2020 (#)
Replying to @ktemkin
Relevant tweet from earlier today (about something slightly different, but definitely universally true, and seems apppropriate here):
https://twitter.com/pjf/status/1332840653030531072
Mon Nov 30 05:22:21 +0000 2020 (#)
TL;DR on AWS issue last week: expanding Kinesis front end pool exceeded OS threading limits trying to mesh, CloudWatch Logs/Metrics use Kinesis, Lambdas push metrics via CloudWatch on invocation (with some buffering, which filled up and impacted Lambdas starting). https://twitter.com/fanf/status/1332983419882643458
Mon Nov 30 05:24:55 +0000 2020 (#)
Replying to @ewenmcneill
It’s a good event retrospective write up, and well worth reading in full, even if you don’t use AWS.
I’d guess until last week hardly anyone would have expected adding Kinesis front end capacity would impact AWS Lambdas (and thus IoT device calls up to the cloud).
Mon Nov 30 05:28:03 +0000 2020 (#)
Replying to @ewenmcneill
And to complete the Fundamental Interconnectedness of All Things, the health dashboard relies on Cognito which also relies on Kinesis. So they had to invent a workaround to publish dashboard updates.
Hosting public status pages on your own infrastructure is an anti-pattern.