Yesterday night, I started autobuilding the archive on mips. I sorted the sources files by age and started with the packages uploaded longest ago. I mainly have two aims: First, like others, I want to autobuild the archive in order to see whether packages still build. Second, from a QA perspective, I'm mostly interested in packages which have not been updated for a while, and I thought autobuilding them would give me a good excuse to look at them in more detail. Over night, about 650 packages were built, and I reported about 10 build failures (mostly out-of-date config.guess files from 2001 which no longer work on mips).
The more interesting part, however, is the sheer number of crappy packages. I know there is a high number of badly maintained packages in Debian, but I was (again) surprised to see how many. There were easily 10-15 packages which said "initial upload", i.e. their last upload was their first upload. First uploaded in 2002, and no single upload since then. Incidentally, in most cases it seems upstream is inactive too (And no, this is not a good reason for not uploading new versions of the package since Debian Policy changes all the time). I really wonder if we had to have those packages in the archive in the first place, and I'll investigate to find out if some of them can be removed. (Fortunately, some have never been part of a stable release, and its still time to get them removed before sarge becomes the next stable release.)
Also, I (again) realized that it is very hard to automate certain QA functions because the right information is not easily available. However, I finally got annoyed enough about the current state to actually do something about it. I will therefore implement a system which will record important data in a way that allows us to automate more QA activities like finding out-of-date or badly maintained packages. More on this on -qa soon.