On Thu, Jan 19, 2017 at 12:57 PM, Junio C Hamano <gitster@xxxxxxxxx> wrote: > Stefan Beller <sbeller@xxxxxxxxxx> writes: > >> git-convert-objects, originally named git-convert-cache was used in >> early 2005 to convert to a new repository format, e.g. adding an author >> date. > > I think this description is not wrong per-se but misses the much > more important point. In the very early days of Git, the objects > were named after SHA-1 of deflated loose object representation, > which meant that tweak in zlib or change of compression level would > give the same object different names X-<. This program was to > convert an ancient history with these objects and rewrite them to > match the new object naming scheme where the name comes from a hash > of the inflated representation. ok, in case I reroll again, I'll fixup the message. > >> By now the need for conversion of the very early repositories is less >> relevant, we no longer need to keep it in contrib; remove it. > > I am not sure if removal of it matters, and I suspect that we saw no > reaction from anybody because nobody thought it deserves the > brain-cycle to decide whether to remove it. I dunno. I do think removing this would improve contrib/, not just because it would better align with contribs mission statement in its README, but also for other reasons. Why would a user look into contrib/ at all? * to find interesting contemporary bits and pieces * if they want to find old stuff for educational purposes, they ought to be looking into contrib/examples instead. So maybe instead of this patch, just move it to the examples section? (That way we archive the same goal: a cleaner, fresher contrib/ that doesn't look as stale) Thanks, Stefan