On Wed, 2008-08-06 at 11:23 +0200, [EMAIL PROTECTED] wrote: > Grant Smith schrieb: > > I was idling in #asfinfra on freenode yesterday and looking at the > > logs now I see they did clean out that repository... > Just to let people know, there is currently some discussion going on at > the infrastructure list about this. > > It was indeed the infra team who deleted all files older than 30 days, > as people.apache.org was running short of disk space. > > The current opinion of the infra admins seems to be that they reserve > the right to delete stuff from the snapshot repo at any time, and that > any reliance on files staying in the snapshot repo for more than a > couple of days is wrong. In particular, some people are concerned that > allowing non-apache people access to the snapshot repo for any purpose > other than testing of artifacts is equivalent to bypassing the release > process, ie any use of this repo except for internal apache development > purposes is wrong (my phrasing). > > A couple of contrary opinions have been expressed; I'll let you know > what the final conclusion is.
Sigh. It seems that everything in Apache descends into a flamewar these days. Anyway, the final conclusion seems to be that infrastructure intends to keep purging all files older than 30 days on a regular basis, in order to keep disk space down. No alternatives or arguments will be considered. And several Apache board members who are also on the infra list back this date-based purging as a way to deliberately ensure that it is hard for any project to publish "stable" artifacts in the snapshot repo that other projects can then depend on, but which were never voted on. If some other project wants to use apache code, it must: * depend only on released artifacts, or * build their own version from source * (or they can copy a snapshot artifact into their own repo) Only one person appears to have any interest in the fact that we currently have artifacts in the snapshot repo that can be unreleased for long periods (or permanently) and which we only rebuild when necessary, eg maven2 plugins or the site skin. So we will need to make some changes here in our dev processes to avoid being broken every 30 days. I think the best solution for us (and every other maven project) is to ensure that we have an automatic build system that builds every artifact in the myfaces family at least once per week, regardless of whether it changed or not. This includes thinks like the site skin, the maven2 plugins and the master pom. This does seem to be some wasted cpu cycles, but it makes sure that the timestamp in the snapshot repo never gets older than 30 days. It does have the nice side-effect of showing a new timestamp when users browse the snapshot repo. I'll try to revive our continuum setup and get this set up. However I know nothing about Continuum, have never administered solaris (or a solaris zone), don't have a login for the zone, etc. So this may take a while. If anyone wants to help (or better, do it themselves..) that would be great. Regards, Simon
