So am I getting this right that the age-old debate about #TIFF vs #JPEG2000 for master images has been officially re-opened at #ipres2026? Like it's 2010 all over again! Or perhaps it never stopped?
Remote
Johan van der Knijff
@bitsgalore@digipres.club
hometown 4.6.9+hometown-1.3.0Digital plumber. Digital preservation, file formats.
526 Followers
265 Following
40 Posts
Joined March 03, 2018
Cool, over 18 years ago I worked as a developer/maintainer of a hydrological simulation model at the European Commission. This is still in active development, and is used for European-level flood forecasting.
Today the current developers (who are not on Fedi AFAIAA ) added the latest source code to #Zenodo, and added #ORCID tags of previous and current contributors, so it's now part of my ORCID profile:
https://orcid.org/0000-0003-3686-337X
If any of them are reading this: thank you, much appreciated!
"The sales pitch for #AI is heavily predicated on the fact that this is all a profitable and sustainable venture that evolves into computerized sentience, and there's simply no evidence that's true.
These models are all absurdly expensive to run, and the mad dash toward pseudo-productivity masked the reality that execs blew gargantuan wads of cash on negligible net benefit and a lot of chronic migraines."
https://karlbode.com/the-ai-hype-reckoning-is-upon-us/
Replying to
@dsalo Here's some useful (I think) context by @davidgerard:
https://pivot-to-ai.com/2026/04/09/claude-mythos-the-ai-hacking-model-too-good-to-release-allegedly/
This also links to this study which basically concludes Mythos' performance is similar to existing, smaller models:
https://aisle.com/blog/ai-cybersecurity-after-mythos-the-jagged-frontier
Interesting, the Bodleian library created a #Python library that converts #TIFF or #JPEG source images to #JP2 #JPEG2000 (using a #Kakadu wrapper), and does various quality checks on the result:
https://github.com/bodleian/image-processing
Will definitely give this a try for an upcoming in-house TIFF to JP2 conversion pilot project!
Oh, Boards Of Canada mail mysterious VHS 📼 tapes to fans, sparking speculation about possible new album:
https://stereogum.com/2495125/boards-of-canada-mail-mysterious-vhs-to-fans/news #BoardsOfCanada #BoC
Replying to
@platypus@glammr.us Another really good ffmpeg resource with lots of adaptable "recipes" is the ffmprovisr website (developed mainly by @ashley@digipres.club):
https://amiaopensource.github.io/ffmprovisr/
Meanwhile on a major digital preservation mailing list I'm subscribed to, someone posted an announcement of a report about "AI for Government Archives".
The announcement e-mail (by the lead author) looks like it's largely #AI-generated, with the typical formatting signs like emoji lists, excessive use of boldface, the works. Haven't looked in any detail at the study itself, but announcing it like this doesn't exactly inspire confidence, to put it mildly 🧐 ...
Useful Firefox add-on that detects and highlights em-dash characters on webpages. Handy for AI #slop spotting:
https://addons.mozilla.org/en-US/firefox/addon/em-dash-detector/
Replying to
@matthewpburgess@digipres.club @anj@digipres.club
In that case use the 1.0.10b8 pre-release, which includes the (mostly untested) fixes. You should be able to install it using something like this:
pip install --pre iromlab
If you could let me know how well that works (if it works at all) that would be really helpful.
Mike Judge still got it - Beavis and Butt-Head Become A.I. Tech Bros:
Beavis and Butt-Head Become A.I. Tech Bros | Animated
Just found out the Open Preservation Foundation (#OPF) has quietly started to embrace generative #AI in its new ODF Validator tool. I have some concerns, so I opened this issue:
https://github.com/openpreserve/odf-validator/issues/293
If anyone wants to weigh in on this, please do!
Replying to
@platypus@glammr.us @ashley@digipres.club Yes it's a brilliant resource. I don't have to process AV files that often, but when I do this is always the first place I look, and it's never failed me so far in coming up with at least a usable starting point!
Maintainer of #chardet, a widely used #Python character encoding detector library, replaces entire existing codebase with #AI-generated code, and changes the license in the process. The original author isn't pleased:
Replying to
"This terminal is linked to a giant brain 10 miles away in the heart of London":
https://doomlode.bandcamp.com/track/a-giant-brain-ten-miles-away-in-the-heart-of-london
Replying to
@Thorsted Yes. There's also Grok (not to be confused with Elon Musk's Grok), which is open source, and had similar performance as Kakadu when I last used it. Details here:
https://bitsgalore.org/2022/03/30/generating-lossy-access-jp2s-from-lossless-preservation-masters.html
Replying to
@BertrandCaron Had a quick peek at the code and found this:
https://github.com/openpreserve/jhove/blob/7595dd5b6660bfe6d2ca9ef3da50ee9df5bebfc4/jhove-modules/jpeg-hul/src/main/java/edu/harvard/hul/ois/jhove/module/JpegModule.java#L1543
So apparently for a 4-component colourspace, JHOVE will either report a value of "CMYK" or "YCCK", depending on the value of "_transformFlag".
Also see (this explains the transform flag):
https://stackoverflow.com/a/50861048/1209004
Replying to
@anj@digipres.club Yes, they're super cheap. You could probably also build one yourself using an old internal drive and some adapter kit, but I've never tried that myself.
Replying to
@anj@digipres.club From what I recall most of these modern drives also don't support 720 kB (DD) disks, which is another good reason to go for refurbished ones.
Replying to
@anj@digipres.club Meanwhile our operational workflow still runs in an old Windows 10 environment.
The imaging/ripping project is almost finished now, and it's unclear when/if there will be an update that fixes these issues (last time I worked on this our IT dept declined to answer my questions about Win 11 behavior, which means development is effectively in a deadlock)
Replying to
@beet_keeper@digipres.club In case of emergency this might help for extracting the table data from the PDF (results are not perfect, but with some editing you'll probably get something usable):
http://tabula.ondata.it/
Replying to
@neil Out of curiosity I ran their main email domain through this tool by @internet_nl, which tests for modern internet standards:
https://internet.nl/mail/thundermail.com/1853398/
Now to be fair I've seen far worse, and the service isn't operational yet. But for a "pro" service it doesn't exactly inspire a lot of confidence (the same tool gives me a 100% score on my own email domain) .
Replying to
@cazabon Not sure if this helps, but the Roland W30 sampling workstation from 1989 (which had a floppy drive) had a demo song called "THE OFFICE" which included some phone and voice samples. A quick search turned up this MP3:
https://www.synthmania.com/Roland%20W-30/Audio/Factory%20Demo%20Disk%20songs/Data%20Disk%202%20-%20THE%20OFFICE%20.1.mp3
Not really an answering machine message though.
Replying to
@Thorsted BTW from what I see it would be really easy to add Grok support to the Bodleian tool. Mostly a matter of creating a copy of the Kakadu wrapper module (here: https://github.com/bodleian/image-processing/blob/master/image_processing/kakadu.py) and then adapt it into a Grok module, and then perhaps adding some imports into the higher-level modules.
Replying to
@BertrandCaron Good question. I have no idea!
Replying to
@BertrandCaron From my (pretty limited) understanding, YCCK is derived from the CMYK color model, similarly to how YCbCr is derived from the RGB model. See also:
https://en.wikipedia.org/wiki/YCbCr
Under JPEG conversion section:
"JPEG also defines a "YCCK" format from Adobe for CMYK input. (...) As a result, a similar set of subsampling techniques can be used"
Replying to
@Thorsted @bishbashbackup @beet_keeper I guess it really depends on the specific use case and foreseen access mode(s). It's pretty easy to paint yourself in a corner with a format that is difficult (or even impossible) to work with in an emulator. (I'm not familiar with libewf myself.)
Replying to
@Thorsted @bishbashbackup @beet_keeper Might not be the best choice if you want to use the disk images in an emulated or virtualized environment though!
Replying to
@Richard_Littler@mastodon.social I would avoid that platform at all costs.
Over the past few months @_elena@mastodon.social has uncovered various major issues with it, see this series of blog posts here:
https://blog.elenarossini.com
Basically it's just a scheme by some politically well-connected people who want to enforce ID verification on social media (they require you to scan and upload your passport) .
Technically it's also on the crap side, so eventually those ID data will most likely end up in a data breach.
Replying to
@matthewpburgess@digipres.club @anj@digipres.club You might also want to check with @kieranjol@digipres.club, who is also about to start some tests with Iromlab on a Win 11 machine.
Replying to
@anj@digipres.club Yes, but keep in mind the Iromlab approach works best for large, diverse collections of optical media that are in comparatively good condition (typically glass mastered discs). E.g. "burned" discs are more likely to give problems, and may need more work (more read passes, possibly using tools that are specifically aimed at data recovery like ddrescue). So it really depends.
Replying to
@matthewpburgess@digipres.club @anj@digipres.club Other thing to keep in mind is that we're still using it with a pretty old (4.x) version of IsoBuster, and I don't know how well it works with the current 5.x releases (the IsoBuster developer gave the software a major overhaul around 5.x). So there might be a few surprises there.
Replying to
@anj@digipres.club And you can still buy them here (they changed the brand name, but technically they're identical):
https://www.conrad.com/en/p/renkforce-rf-4755732-floppy-disk-drive-refurbished-good-usb-a-usb-2-0-2377866.html
Oh, a new Squarepusher track; album to follow in April:
Replying to
@anj@digipres.club Yes. Important caveat: the current "stable" release is incompatible with Python 3.12 and more recent!
I already worked on a fix 2 years ago (see https://github.com/KBNLresearch/iromlab/issues/107) but this is largely untested because Windows 11 blocks the installed executables. Details here:
https://github.com/KBNLresearch/iromlab/issues/113
Not clear if this is a general Windows 11 thing or the result of some internal security policy.
Replying to
Replying to
@anj@digipres.club Never heard about that before!
Might be worth a mention that the USB floppy drives I'm using are refurbished drives (see: https://web.archive.org/web/20181008141513/http://www.produktinfo.conrad.com/datenblaetter/1100000-1199999/001170561-an-01-ml-BASETECH_FLOPPY_LAUFWERK_USB_de_en_fr_nl.pdf)
Perhaps 'modern' ones work differently, I have no idea!