| |||||||||||||
Debian proposes dropping most architecturesDebian proposes dropping most architecturesPosted Mar 14, 2005 22:43 UTC (Mon) by dlang (✭ supporter ✭, #313)Parent article: Debian proposes dropping most architectures
avoiding the coordinated release problem is a good idea if it helps get things out on a reasonable schedule.
however the criteria for becoming a 'top-tier' arch is unreasonably narrow (10% of downloads will be hard for anything to hit, given how much of the downloads x86 is going to be for example)
also this seems to say that second class citizens will never have an official release, only a snapshot (and that on some undefined schedule). this is about as bad as never having a release at all.
it would be far better to have the SCC's not be release showstoppers for the primary arch's. but have the goal that within X months of a stable release on the primary arch's the SCC's will be released (hopefully with the same set of packages, just including fixes that are needed to make them work on that arch, with those fixes propogating to the main tree so that the primaries will get package updates as needed)
(Log in to post comments)
Debian proposes dropping most architectures Posted Mar 14, 2005 23:40 UTC (Mon) by JoeBuck (subscriber, #2330) [Link] No non-x86 system meets the 10% threshold. Even PowerPC is less than 3%. See Dirk Eddelbuettel's blog for some detailed stats.
Debian proposes dropping most architectures Posted Mar 15, 2005 14:51 UTC (Tue) by djpig (subscriber, #18768) [Link] however the criteria for becoming a 'top-tier' arch is unreasonably narrow (10% of downloads will be hard for anything to hit, given how much of the downloads x86 is going to be for example) Note that this criteria is only about mirroring location not about releases. The plan was to release about four arches but only distribute the most needed ones over ftp.debian.org so that mirrors can easier decide to only distribute them. Other release architectures (like PPC) should then be distributed over scc.debian.org (or whatever name would be choosen) which most of the big mirrors will distribute anyway. This should make the live easier for mirrors that don't have the discspace and bandwith for the current archive (>100 GB discspace and 1-2.5 GB daily pulse)
|
|||||||||||||