More interesting use of my time than being in front of a computer.
Remote
Even Rouault
@EvenRouault@mastodon.social
mastodon 4.8.0-nightly.2026-10-06On strike against inaction to fight climate change, and citizens being left aside from decisions regarding AI !
Previously: Open-source geospatial software developer (GDAL, PROJ, QGIS, MapServer, ...). Creator of https://sozip.org
My opinions reflect the ones of my employer, not necessarily mine.
LLMs contribute to global warming and brain rot.
I also think ultra-capitalism should be abolished.
1152 Followers
82 Following
49 Posts
Joined October 28, 2022
0xA th day of strike
7th day of strike. 7 is the smallest whole number whose square root cannot be found as the distance between 2 vertices of a set of contiguous unit cubes.
9th day of strike: your last chance to join the protest before the counter increases to two digits.
The irony is that while being on strike and effectively doing nothing one could argue that I'm still doing something since I've been told that the role of an open source maintainer is to mostly reject contributions (especially in those LLM polluted days)
/me checking we're not April 1st: https://www.phoronix.com/news/Debian-Inference-Portal
8th day of strike. 8 like in octogon, the name of the bar in the student residence where we dreamt of a better world.
Accenture pour aller taper sur les doigts des vilains employés d'Anthropic qui n'arrivent pas à controler leur monstre. Quel cirque... https://www.lefigaro.fr/secteur/high-tech/avoir-un-acces-aux-employes-et-pouvoir-les-embeter-des-consultants-d-accenture-bientot-chez-anthropic-pour-controler-la-securite-de-ses-modeles-20260921?utm_source=firefox-newtab-fr-fr
Petite info utile à quiconque se rend à la déchetterie de Montesson: il est tout à fait possible de répondre "un cadavre" à la question "qu'amenez vous ?" de l'employé à l'accueil, et pouvoir entrer tranquillement.
Au lieu de subventionner les carburants fossiles, ne serait il pas plus pertinent que les finances publiques servent à financer l'achat de véhicules électriques ?
The main skill of a human compared to a LLM ? The ability to say "I don't know" and deal with that inability
If you are a student filing low quality LLM assisted bug reports to one of my projects, I will email your supervisor (I just dit!). Besides that, I'm generally a sympathetic guy
I'd be interested in testimonies of other "semi-maintainers". I consider myself as a "semi-maintainer" of OpenJPEG (maybe "25%-maintainer" would be more accurate), given there is no complementary half/75%. That is I only care about fixing issues that directly affect me/my customers, occasionally merging a "trivial" pull request or doing minimal work to keep CI working. But I'm generally not willing to dig in code I'm not familiar with, including addressing security issues. What to do... ?
#GDAL #Icechunk read-only driver queued at https://github.com/OSGeo/gdal/pull/14755
Japanese geodetic datum JGD2011 just got renamed JGD2024. It is the "same" (same code, etc). What could go wrong ? Maybe just call it "JGDNOW", or "JGDCURRENT" or "JGDWHOCARES" to be really future proof
Let's see if bots get the message: https://github.com/OSGeo/gdal/pull/14612
RE: https://mastodon.social/@gdal/116353864267753547
It is so great to have community members that take with so much serious the challenge of georeferencing the Hello World picture from Artemis II. A promise is a promise: GDAL 3.13 will add that georeferencing automatically 🥳
Wondering if I'm not breaching code of conduct writing in a OGC venue that something is over-engineered 😂
https://github.com/opengeospatial/geotiff/issues/56#issuecomment-4496478301
Nothing better than preparing my #FOSS4GE2026 workshop to file new fix/improvement issues
#dogfooding
Anyone knows which consortium has won the HORIZON-CL6-2026-03 tender (6 million euros for Open-source geospatial software)?
https://ec.europa.eu/info/funding-tenders/opportunities/portal/screen/opportunities/topic-details/HORIZON-CL6-2026-03-GOVERNANCE-06
We officially got our first vibe coded pull request where the bot proudly identified itself with ridiculous emoji as instructed in CONTRIBUTING.md . Deep sigh... What a fucked up world.
Not promising at all to satisfy dreams, but what is the GDAL driver that you have dreamed about and that has never materialized? (humoristic answers accepted)
Is GDAL really a big beast ? It can, but you can also tame it:
- Minimal build with only Zarr, Icechunk and VSI support: libgdal.so.39.3.14.0: 14.6 MB
- vs equivalent Python/Rust stack "du venv/lib/python3.12/site-packages/icechunk venv/lib/python3.12/site-packages/zarr venv/lib/python3.12/site-packages/numcodecs": 87.3 MB
AI coding agents try so many sub-commands (rg, grep, sed, awk, etc.) that apparently the fork()+exec() paradigm to spawn processes is not fast enough when they need hundreds of spawns per second! That said the attempt at https://lwn.net/ml/all/20260528095235.2491226-1-me@linux.beauty/ to improve that isn't very convincing. A mere 3% improvement
One the few advantages of software security being mostly AI detection driven nowadays is that there is no more any reason to treat vulnerability reports privately, including reproducer. If someone reports to you an issue, 100 other people have found it already.
Replying to
Ceci dit assez intéressant à écouter car relativement peu de langue de bois. Sa vision est que le développement logiciel est transformé de l'artisanat vers celui de manager d'agents. Effectivement, je dis souvent que je me vois plus comme un artisan! Ai-je envie de devenir le manager de robots non déterministes et dont aucun humain ne peut expliquer le fonctionnement... Je préfère largement que GCC m'insulte à cause de mes templates erronés ! Au moins il ne me dit pas "vous avez raison"
Replying to
https://github.com/OSGeo/gdal/pull/14749 should make GDAL users' life easier (provided the run it from us-west-2 EC2 instances...)
Replying to
@ianturton@mapstodon.space did someone tell EPSG that FOSS4G Europe was in Europe this year and they'd better not mess with ETRS89 😛
What am I doing ?
dd if=test-dataset/repo of=/dev/stdout bs=1 skip=39 | zstdcat -d > repo.bin && flatc --raw-binary --json repo.fbs -- repo.bin && cat repo.json
The new "gdal" command line utility that will make its prime appearance in GDAL 3.11 has now a quite substantial number of commands ported to it: https://gdal.org/en/latest/programs/index.html#gdal-application
17 pull requests submitted by one "person" in a hour to GitLab (by a GitLab employee). What can go wrong ?
Replying to
@jakobmiksch Likely, but always ask for consent beforehand. Realizing you've served as a guinea's pig to someone else social experiment always live a bad taste. Anyway the outcome is known: if you try hard enough, you'll succeed in introducing bugs not found during review. Review is not about preventing all bugs to be introduced but catching at least a few of them..
Replying to
@ianturton@mapstodon.space there is a constant tension between creating new versions of objects to reflect an updated geodesic reality and not overwhelming users that don't care about those subtelties with an avalanche of new codes. No easy solution to please everybody
Replying to
@neteler@mastodon.social @mundialis@mastodon.social @jjimenezshaw@mapstodon.space Der Titel des zweiten Bildes lautet „Transversale Mercator ist dein Freund“, richtig?
Replying to
@jakobmiksch@mastodon.social I love this sentence in https://www.scirp.org/journal/paperinformation?paperid=137296 : "Java code often includes more boilerplate and verbosity". It is refreshing to read things written with such honesty 😂
Replying to
@bagder I hope you get big $$$ from those corporations to participate to their advertising campaign...
Replying to
Replying to
@JensWiesehahn@fosstodon.org @MikeMahoney218@fosstodon.org GDAL docs are already a 1,200 page PDF (excluding C/C++/Python/Java API documentation). They are admittedly more a reference manual than a tutorial one. The majority of features are documented, but that might be just one or two terse sentences to uncover. Tutorial type of docs would be likely like 5 or 10K pages. That said, we are always open to sensible doc enhancement contributions.
Replying to
@JensWiesehahn@fosstodon.org @MikeMahoney218@fosstodon.org 15+ years users of GDAL can still discover features they weren't aware of: https://github.com/OSGeo/gdal/issues/11100#issuecomment-2433137832 . Well, 'find alg gcore gnm frmts port ogr apps swig \( -name "*.cpp" -o -name "*.c" \) -exec cat {} \; | wc -l' reports 1.7 million lines, so no wonder there are a few hidden gems
Replying to
@aljazmerzen@fosstodon.org Hack or rather... Easter Egg :-) But how could we do better ? The EXIF parameters or other metadata included in the image provided by NASA aren't enough.
Replying to
@b@mapstodon.space the pace of human bad quality is slower and we are used to deal with that. We just cannot cope with the pace of machines. This is an existential fight.
Replying to
@aljazmerzen@fosstodon.org Too much work for an easter egg. Checking image filename + dimensions is enough IMHO to be reasonably confident it is the one we expect and avoid false positives.
Replying to
@erilem@fosstodon.org c'est un problème que j'ai en effet en tant que mainteneur face à une contribution. Il m'arrive souvent de dire à un contributeur "ne pourrait on pas plutôt faire comme ça ?" Ma suggestion est parfois erronée et un humain avant d'y plonger à deux pieds va éventuellement la réfuter. Le LLM finira par la faire marcher.
Replying to
@implgeo@mapstodon.space ah didn't remember this was one of the killer feature :-) Should it be mentioned in the "in a nutshell" section ?
Replying to
@bagder It was probably badly phrased of mine. I didn't intend to be rude. It was more a reflexion that us open source devs are indirectly serving as marketing material for those black box highly capitalistic products.
Replying to
@ctoney @jakobmiksch Thanks Chris for finding back this occurence. I knew this was a déjà-vu