IRC channel logs

<ArneBab_>old: restack includes the goal of a more resilient stack. With Guix, reproducibility via Guile is great. From the page: "In scope are innovations and enhancements to core internet infrastructure, alternatives to closed online technologies serving a large userbase, and devices used to run and consume these in a secure and privacy-friendly manner. This also includes full stack security, software supply chain management, reproducibility and

<ArneBab_>trust enhancing technologies."

<ArneBab_>old: examples that could be a good match: harden (fibers web), build more networking protocols, simplify the stack (no need for many packages to have a working system), spritely hoot simplifies the stack (only need to know guile scheme), enhancing guix graph with sbom output to be able to check CVEs (to try: guix graph -b cyclonedx-json perl), spritely goblins is a perfect match for trust enhancing technologies.

<ArneBab_>old: also adding a linter that helps avoid common pitfalls – both in security and in performance. My BSI Grundschutz site documents several missing pieces to improve the default security of applications based on Guile.

<ArneBab_>old: and why I post it here: people can get 5.000 to 50.000 Euro grants for such projects.

<ArneBab_>⇒ I want people to be paid for working on Guile.

<old>ArneBab_: I would like very much to be paid for that :-)

<lilyp>sneek: botsnack

<sneek>:)

<tbennett>Hello, I'm having a hard time understanding how to use SXML library's pre-post-order function. Do you know where I can look for examples?

<ray1729_>tbennett: I only used it once - https://codeberg.org/guix/planet-guix/src/branch/main/planet/parse/common.scm#L65

<tbennett>@ray1729 oh thank you! It's the handler arguments that I didn't get.

<sneek>Yey! dsmith is back

<dsmith>sneek, botsnack

<sneek>:)

<ArneBab_>old: just look at https://nlnet.nl/restack/ and https://nlnet.nl/restack/eligibility/ to see whether your ideas match. If yes, then start in https://nlnet.nl/restack/guideforapplicants/ – that’s 5 pages to read to understand the process. Then write at most 18.000 characters of text (in total) in the proposal form: https://nlnet.nl/propose/

<ArneBab_>old: I’ve done three projects for nlnet; two successful, one unfinished due to family health reasons. The amount of bureaucracy is nicely low.

<JohnCowan>those directions sound disturbingly like a prompt

<old>ArneBab_: right I did asked for a grant under another nlnet project for BLUE

<jlicht>old: did you get it?

<old>jlicht: nope

<ArneBab_>JohnCowan: no prompt, just the shortest way I could write the guide (I looked up the URLs to get rid of any potential hurdle)

<ArneBab_>JohnCowan: but reading it, I understand why it seems like one … (but: it would totally fail, because those 18k characters to write aren’t *one* text, it’s just the sum of the limits given in the forms)

<ArneBab_>… correction: it would likely work ☹ please don’t use it that way. I think that using an LLM to generate the proposal would be a betrayal of the intentionally low bureaucracy at nlnet.

<ekaitz>ArneBab_: i don't think nlnet would be happy with llm generated work

<ekaitz>in fact I think they reject it

<old>I pushed the DCO and Coding assistant in HACKING. LLM is stricly ban for any contributions made to Guile

<ieure>old, Love to see it. Thank you.

<dthompson>old: thank you!

<bremner>old: yay!

<old>I also added a new PR template for CB. Any new PR will have to provide [X] for the following statements:

<old>- [ ] I certify that the code that is contributed in this PR was not generated, either fully or partially by a LLM. - [ ] I certify that I have the right to publish this code under the license of the project and I agree with the Developer Certification of Origin.

<old>I don't know if this will work with Agit tho

<old>and sorry if this can be anoying (to check-box the statement everytime), but I think it's a necessary evil

<JohnCowan>old: the second implies the first at present, but that may change

<JohnCowan>nobody can certify that they have the right to publish LLM-generated code, because nobody knows who owns it

<bremner>yes, but that's a debate about a particular legal system

<bremner>so I think being explicit is good

<bremner>(fwiw)

<old>like you said, this is subject to change so I don't think it hurts to certify both

<old>Question

<old>which modules you think is important to be fast to load for user programs

<old>I'm making some standard metrics gathering

<old>to detect performance regression of Guile

<old>for example, I am measuring the time it takes for: guile -c '(use-modules (oop goops))'

<old>same for (rnrs base)

<old>any other module that are heavily used and could impact load time?

<rlb>Agree goops, likely rnrs, though I still don't use those heavily yet, and perhaps also the really common srfis, though I guess a number of those are already in (guile).

<rlb>match?

<rlb>format?

<old>srfi-1 ?

<rlb>That one's already in (guile)?

<old>right

<old>well there is (system vm assembler) and (system base compile) which are very slow

<old>comapre to other

<old>actually compile seems to be very fast

<old>assembler not so much

IRC channel logs

2026-09-09.log