Hi Andreas, On 07-11-2023 18:01, Andreas Tille wrote:
You did not yet answered the question I asked twice whether we can find a compromise by simply removing packages with missing new dependencies from testing. I consider this a compromise which I would really love to discuss honestly.
I'll try to give at least one answer to this question and also add some other remarks after that.
The first and foremost reason why we're not enthusiastic about defaulting to removal of packages from testing is that that's a disservice to our users of testing. Yes, we (auto)remove packages from testing, but we mostly do that to remove RC buggy packages that aren't fixed in time. Removal out of the blue is rare. For the current transition, if we remove packages from testing because they are not ready and the transition migrates, these removed packages will suddenly cause systems where any of these packages is installed to hold back a lot of related packages and the user will have to figure out which one is causing the situation, why and how to solve it for their personal situation. (Were we to remove packages in other situations, still installed packages may suddenly fail to run without a proper bug report explaining the situation. In this case that's prevented by the r-api-bioc relation.)
Apart from the reasons already mentioned by Sebastian, another reason why at least I am not looking forward to an r-bioc-biocgenerics transition is that I have the *feeling* (which might be unjust, but then consider past interactions) that we often need to tell you about issues and how to solve them. If the r-bioc team were to actively monitor the situation and respond to failures and other problems (you might be already doing this, it might just be a visibility problem) then it would be easier on us even if a transition last a bit. As Sebastian said, for most other transitions the period that requires quite some attention from us is shorter, which means less pushing and popping from stack, and thus a lower mental load. We're doing quite a lot of transitions, longer lasting ones are never nice.
Several other ecosystems, e.g. perl, python and ruby, do preparation outside of the Debian archive and often have some idea (or know quite well) of what to expect during the transition. It's that pro-activeness that makes a difference to us; it *sounds* much better than "let's start and find out how much work it is". Maybe it's just the tone, but we're all humans.
I realize that the current tools and way of working in the r-bioc ecosystem isn't yet geared towards our ideal mode of operation. If you could try and work towards that, that would be appreciated.
Paul
Attachment:
OpenPGP_signature.asc
Description: OpenPGP digital signature