#60 Indicate that FESCo is responsible for all decisions relating to whether a service may autostart
Closed: Fixed None Opened 11 years ago by spot.

FESCo is responsible for defining an approved list of services which may autostart, and/or defining criteria to determine when and if a service may autostart.


Here's my list of reasons for voting -1 on this:

The flaw I found with the bundled library exceptions handling by FESCo was that it lead to inconsistent treatment of apps asking for exceptions. If you asked for an exception in F8, you got denied but if you asked for an exception in F12 with pretty much the same rationale you were approved. My feeling is that a body that changes part of its composition every 6 months helps lead to this inconsistency. Inconsistency of application leads to packagers feeling slighted which is not beneficial. A decision to use words that carry no guidance (using "approved" instead of "essential" for instance) also makes this problem worse as there's no baseline understanding of what's being achieved by limiting what can autostart at all.

Decisions of how to package are FPC's to decide. Decisions of what to package are FESCo's. In this case we aren't saying that a particular class of packages may or may not be packaged -- we're attempting to tell how to judge which recipe (autostart or not) to follow for a particular service. My thinking is that this is very much a how to package problem.

The discussion about this seemed to be that if we could come up with criteria then it was within FPC's jurisdiction to approve this. Because we didn't agree on criteria, this was deemed as a FESCo matter. To me the criteria should not be the reason that something is removed from what the FPC decides upon. Rather it's whether we're deciding how to package or what to package. Similar decisions are made within FPC about whether a particular file is an arch-specific data file and goes in %{_libdir} or arch independent and goes in %{_datadir} -- we're just able to agree on the criteria to use in those cases more easily than we have in this case.

The draft guidelines that have been proposed by FESCo have the same issues that caused us to originally disagree on criteria (they are almost verbatim the same criteria). So handing to FESCo doesn't address the original problems with the criteria that caused us disagreements in the first place.

I'm +1 on this. Determining how a package enables a service is FPC's responsibility. Determining which services can be enabled is FESCo's responsibility.

+1, although I'm much closer to "meh." than either spot or toshio :)

So I'm going to explain my rationale, as a middle of the road voter:

  1. I don't think the bundling problem will happen here, mainly because "stuff that starts by default" has a much more visible result than "bundling".

  2. If we decide it, the answer we give can't be different for different spins and ideally I think there are a lot of cases where "default enabled" should be different for different spins (even assuming FESCO can't do that either, right now, it has a much better chance of getting there if they own it).

  3. I think there is a good argument that you can split the "packaging" problem from the "default on" problem, and it's only obvious we should own the former.

  4. In general it seems better to let FESCO (the elected body) handle things like this (where everyone wants to be on the list, but only a small number should be and who gets on it is very subjective), at least until that's proven to be a problem. This should make all the packagers happier, even when they get the same result.

Decisions of how to package are FPC's to decide.


Replying to [comment:3 rdieter]:


I'm ambivalent about this, and I can see arguments either way. However, what sways me more towards "-1" is that we're obviously responsible for writing the packaging guidelines, and I can imagine that fesco would have problems with being responsible for granting exceptions but not for crafting the rules for granting those exceptions (which to me would belong in the guidelines).

So rather than abdicate this entirely and say "fesco is responsible for this portion of the guidelines" I'd rather just do it all. We already do this for bundled library stuff with good reason, and I really don't see how this situation is significantly different.

I wouldn't really complain if it went the other way, of course, since that appeals to my sense of doing less work and the amount of work we do seems to be ramping up lately.

-1 from me as well. Although pushing the responsibility on to FESCo is tempting, I see no reason why we should be less qualified to produce both the rules and the exception list. In my opinion, this falls into "packaging guidelines" category.

+1 From me, because I consider "which services to start" to be a mostly political and distribution strategic decision => FPC's responsibility.

Login to comment on this ticket.