IRC channel logs


back to list of logs

<nckx>Would reopening the tarball vs. git (and tag vs. commit) discussion on an ML be a good idea, or just annoying? I'm in favour of both latters myself, but it seems hard to forge consensus on this.
<nckx>And annoying folks will just lead to digging in.
<civodul>nckx: i think both are worth discussing, but both are tricky
<civodul>in both cases one needs to be aware of the archival implications
<civodul>for tarball vs. git there are bootstrapping considerations
<civodul>i remember a discussion where we had reached some sort of a consensus on the way forward
<nckx>Right. I mean the general (vast majority) case, not to be pedantic about reasonable exceptions.
<nckx>civodul: Would you happen to have the link? I failed to find it.
<nckx>Or just paraphrase it?
<civodul>nckx: i don't have a link; i think Maxime Devos started the thread
<nckx>To finish the previous thought: with a smaller set of bootstrap (or other) packages that do use tarballs, we can actually improve our availability guarantees, mirror them, back them up, etc.
<nckx>OK, thanks. I'll search again at home.
<civodul>re backup it's not clear-cut now that we have Disarchive
<nckx>Does disarchive produce bit-identical output?
<civodul>yup, that's its raison d'être, if i may sound fancy
<civodul>we never got a real blog post about it unfortunately (because it's really nice & fun) but there are bits at
<nckx>Fancy? You just sound French.
<civodul>French words in English sound fancy, no? :-)
<nckx>I knew what disarchive ‘did’, but not that it did so that well. Silly me.
<civodul>we're not at 100% coverage, but it's reached the point where tar.{gz,xz} archival is much less of a concern
<nckx>Less of a concern than movable git tags, then.
<nckx>Does Disarchive just reject a file if it was compressed with a format-compatible but not bit-identical compressor?
<nckx>I guess this is veering off-topic.
<civodul>yes, Git tags are a concern
<civodul>yes, Disarchive bails out if it fails to find the right compressor flags