Pascal Quach

Just trying to install guix system but my wifi card is not supported hardware 😢 and I don't have ethernet

#guix

Ever since my PhD contract has ended last month, I have been going through my personal project backlog, including this very Mastodon account and actually making progress on my personal website.

Yesterday, I was thinking what to do about my Goodreads account when I discovered @bookwyrm@tech.lgbt and by proxy @OpenLibrary@mastodon.archive.org and @inventaire@mamot.fr ! Obviously, I am now on bookwyrm.social.

Because my Goodreads shelves are a mess, many of my books can't be imported automatically, including some japanese light novels and web serials. The latter, I've never really bothered to track outside of dedicated websites (novelupdates, ao3, fanfiction.net) because of how difficult it is.

Big question for people in the know and #librarians, is there a database of web serials or am I doomed to add my books manually (or programmatically)?

#bookwyrm #openlibrary #inventaireio #askfedi

A few years ago, I tried setting up Guix System on WSL2. I succeeded after going through obscure GitHub issues and gists. Eventually, I stopped tinkering with it due to lack of time, but I'm happy to say that if I wanted to try again it's relatively easy to get a working Guix System on WSL2 with or without a working guix!

Arian Dehghani explains in his blog [2] that Guix provides an image builder for WSL2 such that you can install Guix on WSL2 simply by importing a (suspicious) pre-built tarball [1]!

[1] https://github.com/Arian-D/guix-wsl/
[2] https://arian-d.github.io/blog/posts/guix-on-wsl/

1/2

#guix #WSL

RE: @monterosato@mamot.fr

"Sans qu'on puisse choisir, les sociétés commerciales dont nous sommes devenus dépendants décident désormais que nous allons utiliser des agents conversationnels pour tout faire, même quand leur utilisation n'est pas pertinente, même quand elle est complètement aberrante. Et quelle est notre réaction? Nous disons: et ben, on ne peut pas résister au progrès, il faut s'y faire. "Ces technologies sont là pour rester", il faut former les jeunes à les utiliser."

Une belle description du fatalisme rampant présente autour de l'IA générative. Malheureusement plus le temps passe, plus la probabilité d'être contraint à utiliser ces outils est haute, que ce soit par obligation hiérarchique ou par pression de performance. La prophétie devient réalité si tout le monde en est persuadé...

Concernant cette dernière, il suffit de constater l'utilisation massive de l'IA dans la recherche, pour programmer (le code est un moyen), pour écrire ou même pour évaluer ses pairs.

#NoAI #PublishOrPerish #resistiagen

1. Démonter son vélo électrique devant les quais avant d'embarquer comme indiqué
2. Changement de Hall et se trimbaler 25kg sur l'épaule
3. Arriver devant sa voiture et se rendre compte qu'il reste des places de vélo
4. Pleurer

😭😭😭😭

#velo #sncf

I just discovered that Chronicles of Ancient Darkness (Michelle Paver), apparently had three whole new books added to the series in 2020??

Am I about to re-read an entire book series (about 500k words, not that long all things considered)... maybe.

#books #fantasy #bookstodon

Pour ceux qui n'auraient pas encore lu le billet d'Agnès Magron sur le blog du CCSD [2] (l'équipe de @hal_fr@social.numerique.gouv.fr), c'est une lecture importante car elle met en évidence les conséquences de la vague d'IA générative qui déferle sur le monde ...

Je cite "Le volume enregistré en 2025 est deux fois plus important qu’en 2024. [...] Pour donner l’exemple de la journée du mardi 17 février, sur un total de 17 millions de requêtes, ce sont plus de 4 millions de requêtes de robots qui ont été interceptées".

En 2026, HAL compte tester Anubis [1], un des utilitaires majeurs (en termes de communauté et d'efficacité) pour bloquer le traffic de "mauvais" robots. Cette solution est déjà utilisée à travers le monde par des organisations réputées qui n'avaient sûrement pas d'autres choix: l'UNESCO, la fondation Linux, Gnome pour en citer quelques unes.

Cela peut donner lieu à des images un peu cocasse car le branding de @techaro.lol est un poil mignon dans un style #anime !

cf. screenshot du logo de la République française avec la mascotte Anubis sur le Gitlab de l'INRIA

[1] https://en.wikipedia.org/wiki/Anubis_(software)
[2] https://www.ccsd.cnrs.fr/2026/02/bad-bots-good-bots-filtrer-les-robots-pour-reduire-leurs-impacts/

I dabbled a little bit in #lean and found the "tactic" concept to be a very intuitive interface to types metaprogramming.
I'm currently tinkering with my blog generated by #hakyll and as usual the hardest in programming is not writing the code, but discovering interfaces.
Given how good the typing is in #haskell, I thought... Hang on, I'm writing all these definitions, surely the LSP could be smart enough to just let me discover the relevant functions right? Something like a tactic but for Haskell.
One search after, I'm discovering Wingman [1], a tactic metaprogramming plugin for HLS which used to work back in GHC<9. Apparently the upgrade cost was too high and the plugin got dropped unfortunately...

[1] https://reasonablypolymorphic.com/blog/towards-tactics/

As is the case for nearly all complex systems, changing a process is going to be not only difficult, take a long time but also meet a "chicken & egg" problem (causality dilemma). The goal of science is to create and disseminate knowledge. At least for publicly funded research projects, the deliverables should be publicly available. In the past decade, much progress has been made to reach #openscience, including moving from a closed journals to an "open-access" model (whether it actually reduced the financial burden on organizations is another matter entirely).

In the case of the open science reform, the "chicken & egg" problem asks whether researchers should practice open science first, or should the research ecosystem adopt open science first?
1) With strict resource constraints (time, funding, workforce), researchers don't have time to dedicate to deliverables that don't provide a direct value. The values of open science are laudable, but they are not rewarded. Why would researchers spend time creating and sharing a FAIR dataset when publishing another paper with that time is preferable for career advancement?
2) If the research ecosystem adopts open science (one way or another), e.g., researchers MUST publish code, data, register protocols, etc., then there is a huge non-adoption risk, or even worse a malicious compliance risk. A highly valued right for researchers is their academic freedom. In some ways, project-based funding is already threatening academic freedom, then what would it look like if grant agencies started withholding further funding because they didn't publish X or Y dataset?

In an era and a society where work is often meaningless or people are not convinced by their company's values and purposes, research is probably one of the remaining job where the work is driven by the person themselves. For that reason, the culture is very strong and/or opinionated because people actually like their jobs.

Hidden in the problem above is the interdependency between culture, trust and research reforms. Jonny Coates points many issues in a thoughtful piece [1], including that researchers MUST be at the center of the system reform: "Many reform initiatives are designed for researchers, but not often with them. If open science, assessment reform, and research culture initiatives are to succeed, active researchers need to be more than subjects. They need to be participants in designing the systems they are expected to use.".

[1] https://upstream.force11.org/connecting-reform-movements-linking-research-culture-trust-and-open-research-publishing/

#OpenScience #CoARA #ReformingRA

J'ai l'agréable surprise de découvrir que Lime a introduit un nouveau modèle de vélo avec un siège bébé ! À voir si l'adoption suit...

#velo #paris

https://www.li.me/fr-fr/blog/parents-lime-d%C3%A9ploie-500-v%C3%A9los-%C3%A9quip%C3%A9s-de-si%C3%A8ges-enfants-%C3%A0-paris-pour-faciliter-vos-trajets-du-quotidien

@cameronneylon I don't think there is an "accepted standard". Probably the most generic you could use is Dublin Core.

There is a XMP dc namespace https://www.exiftool.org/TagNames/XMP.html#dc which could be used to add metadata to the image directly.

Something that can survive passage through a presentation an added bonus!

I wouldn't trust any processing steps to preserve image metadata though. I think most softwares just strip an image clean before using it.