EPEL vs FESCo stalled process #347

Open
opened 2025-08-28 06:11:29 +00:00 by carlwgeorge · 11 comments
Owner

FESCo has added a stalled request process as an alternative to the full non-responsive maintainer process. This was based on the EPEL stalled request process. Now that it exists, we need to decide if we still need the EPEL process, or if we should align with Fedora and use the FESCo process for EPEL stalled package requests. There are some key differences worth noting.

EPEL process

  • two weeks period with the needinfo flag set
  • no explicit approval, just meet criteria and releng processes it

FESCo process

  • one week period with the needinfo flag set
  • requires an approved FESCo issue

If we deprecate the EPEL process and point people at the Fedora process, that means packagers have to file two tickets (FESCo + releng) during the process instead of one. Maybe this is fine and it doesn't make sense to have two similar processes, or maybe we want to keep the EPEL process to avoid the FESCo issue requirement.

Another aspect is the fact that the FESCo process is described as a "trial process". What does this actually mean? Is the policy more open to changes during the trial? What is the likelihood the process being discarded after the trial? How long is the trial?

@salimma As the proposer and author of the FESCo policy, it would be awesome if you could weigh in on this. I think I remember the original intent of the FESCo policy was for us to be able to not have a separate policy for EPEL, but I could be misremembering.

FESCo has added a [stalled request process](https://docs.fedoraproject.org/en-US/fesco/Policy_for_nonresponsive_package_maintainers/#stalled-request-process) as an alternative to the full non-responsive maintainer process. This was based on the [EPEL stalled request process](https://docs.fedoraproject.org/en-US/epel/epel-package-request/#stalled_epel_requests). Now that it exists, we need to decide if we still need the EPEL process, or if we should align with Fedora and use the FESCo process for EPEL stalled package requests. There are some key differences worth noting. ### EPEL process * two weeks period with the needinfo flag set * no explicit approval, just meet criteria and releng processes it ### FESCo process * one week period with the needinfo flag set * requires an approved FESCo issue If we deprecate the EPEL process and point people at the Fedora process, that means packagers have to file two tickets (FESCo + releng) during the process instead of one. Maybe this is fine and it doesn't make sense to have two similar processes, or maybe we want to keep the EPEL process to avoid the FESCo issue requirement. Another aspect is the fact that the FESCo process is described as a "trial process". What does this actually mean? Is the policy more open to changes during the trial? What is the likelihood the process being discarded after the trial? How long is the trial? @salimma As the proposer and author of the FESCo policy, it would be awesome if you could weigh in on this. I think I remember the original intent of the FESCo policy was for us to be able to not have a separate policy for EPEL, but I could be misremembering.
Contributor

As it currently stands, I like our EPEL stalled process procedure better, and unless they change the FESCo process, I say we do NOT remove ours.

Reasons Why:

  • The longer two week period for the second request (making 3 weeks in total).
    -- We originally had just one week and maintainers complained because it was possible that they were on vacation for 2 weeks and came back to find someone had taken their package. Occasionally there was legitimate reasons why the package shouldn't go in EPEL.
  • Requesting an approval from FESCo:
    -- Technically EPEL falls under FESCo. But why have an EPEL Steering Committee if everything goes through FESCo.
    -- I don't really want it going to the EPEL Steering Committee either. That just adds more time to the ticket, and more burdon to the Committee.
  • Documentation:
    -- Our documentation has examples and is much more geared towards EPEL.
    -- Some of their documentation of what is needed before you do the request is better. I don't mind grabbing some of that.

There is also the matter that their process hasn't been "broken in". But even if we give it 6 months and several processes going through, if their process doesn't change, then my above arguments would still stand and I would not want to change.

As it currently stands, I like our EPEL stalled process procedure better, and unless they change the FESCo process, I say we do NOT remove ours. Reasons Why: * The longer two week period for the second request (making 3 weeks in total). -- We originally had just one week and maintainers complained because it was possible that they were on vacation for 2 weeks and came back to find someone had taken their package. Occasionally there was legitimate reasons why the package shouldn't go in EPEL. * Requesting an approval from FESCo: -- Technically EPEL falls under FESCo. But why have an EPEL Steering Committee if everything goes through FESCo. -- I don't really want it going to the EPEL Steering Committee either. That just adds more time to the ticket, and more burdon to the Committee. * Documentation: -- Our documentation has examples and is much more geared towards EPEL. -- Some of their documentation of what is needed before you do the request is better. I don't mind grabbing some of that. There is also the matter that their process hasn't been "broken in". But even if we give it 6 months and several processes going through, if their process doesn't change, then my above arguments would still stand and I would not want to change.
Contributor

I agree with Troy on pretty much everything here.

I agree with Troy on pretty much everything here.
Contributor

At least for the time being, I'd rather leave things as-is.

At least for the time being, I'd rather leave things as-is.
Author
Owner

The longer two week period for the second request (making 3 weeks in total).

In practice I think the FESCo process will usually also work out to about 3 weeks total, as the FESCo ticket voting period is one week unless fasttracked. As an aside, I do like that their policy has a fasttrack option, so that would be a benefit of standardizing on the same policy.

I don't really want it going to the EPEL Steering Committee either. That just adds more time to the ticket, and more burdon to the Committee.

I mostly agree. I like that our policy is just "meet criteria". I don't think the FESCo vote requirement is overly burdensome, but I do think it's unnecessary. If this this ends up being the main sticking point on this becoming a shared policy, then we should see if FESCo is willing to drop that part.

Documentation:

I think the docs are a wash. Whatever we eventually decide on, we'll tweak the docs to match. That may be separate policies with docs to match, or a shared policy with docs that call out any EPEL differences as needed.

> The longer two week period for the second request (making 3 weeks in total). In practice I think the FESCo process will usually also work out to about 3 weeks total, as the [FESCo ticket voting period](https://docs.fedoraproject.org/en-US/fesco/#ticket-votes) is one week unless fasttracked. As an aside, I do like that their policy has a fasttrack option, so that would be a benefit of standardizing on the same policy. > I don't really want it going to the EPEL Steering Committee either. That just adds more time to the ticket, and more burdon to the Committee. I mostly agree. I like that our policy is just "meet criteria". I don't think the FESCo vote requirement is overly burdensome, but I do think it's unnecessary. If this this ends up being the main sticking point on this becoming a shared policy, then we should see if FESCo is willing to drop that part. > Documentation: I think the docs are a wash. Whatever we eventually decide on, we'll tweak the docs to match. That may be separate policies with docs to match, or a shared policy with docs that call out any EPEL differences as needed.
Contributor

I agree with Troy. I think we should leave things as they are.

I agree with Troy. I think we should leave things as they are.
Author
Owner

Metadata Update from @carlwgeorge:

  • Issue untagged with: meeting
**Metadata Update from @carlwgeorge**: - Issue **un**tagged with: meeting
Author
Owner

We discussed this at today's EPEL Steering Committee meeting.

I'll file a pull request to determine if FESCo is willing to drop the explicit vote portion of their process. After that is determined, ESC will revisit this.

We discussed this at today's EPEL Steering Committee meeting. I'll file a pull request to determine if FESCo is willing to drop the explicit vote portion of their process. After that is determined, ESC will revisit this.
Author
Owner
https://pagure.io/fesco/fesco-docs/pull-request/110
Contributor

Finally back from AFK, sorry, relocation took forever.

For context

  • I initially made the trial FESCo process longer (similar to EPSco) and... was told that's too long. It's hard finding a balance
  • The intention is to deprecate the EPSco process eventually, but I'd rather have this tested first before dropping the EPSco one
  • I think if this works for FESCo (regardless of whether it works for EPSco too) we would likely want to revisit if the nuclear option (of orphaning everything a non-responsive maintainer holds) should be used. So something will change - either the trial process does not work, and need revising, or it works really well and we can reduce the scope of the full-blown process and direct people to the faster more lightweight process unless the maintainer in question is seriously non responsive.
Finally back from AFK, sorry, relocation took forever. For context - I initially made the trial FESCo process longer (similar to EPSco) and... was told that's too long. It's hard finding a balance - The intention is to deprecate the EPSco process eventually, but I'd rather have this tested first before dropping the EPSco one - I think if this works for FESCo (regardless of whether it works for EPSco too) we would likely want to revisit if the nuclear option (of orphaning everything a non-responsive maintainer holds) should be used. So *something* will change - either the trial process does not work, and need revising, or it works really well and we can reduce the scope of the full-blown process and direct people to the faster more lightweight process unless the maintainer in question is seriously non responsive.
Author
Owner

Thanks for chiming in @salimma. That is pretty much what I remembered/expected.

I'll wait for feedback from FESCo on that docs PR proposal to drop the voting step.

In one of the ESC meetings I also agreed that I would take a package through the FESCo process to get a feel for it. I just don't have a candidate package for that yet.

Thanks for chiming in @salimma. That is pretty much what I remembered/expected. I'll wait for feedback from FESCo on that docs PR proposal to drop the voting step. In one of the ESC meetings I also agreed that I would take a package through the FESCo process to get a feel for it. I just don't have a candidate package for that yet.
Author
Owner

The conclusion on fesco#110 was that FESCo still wants some approval, but is willing to relax it to just a single FESCo member ack in the releng ticket, avoiding the need for two separate tickets. I was going to adjust the PR to say that, but @zbyszek beat me to it and pulled my original commit into fesco#114 with an additional commit for the since ack change. That has been merged, so the next step is to use this process and see how it feels, then we can discuss if the separate EPEL process is still necessary.

The conclusion on [fesco#110](https://pagure.io/fesco/fesco-docs/pull-request/110) was that FESCo still wants some approval, but is willing to relax it to just a single FESCo member ack in the releng ticket, avoiding the need for two separate tickets. I was going to adjust the PR to say that, but @zbyszek beat me to it and pulled my original commit into [fesco#114](https://pagure.io/fesco/fesco-docs/pull-request/114) with an additional commit for the since ack change. That has been merged, so the next step is to use this process and see how it feels, then we can discuss if the separate EPEL process is still necessary.
Sign in to join this conversation.
No milestone
No project
No assignees
6 participants
Notifications
Due date
The due date is invalid or out of range. Please use the format "yyyy-mm-dd".

No due date set.

Dependencies

No dependencies set.

Reference
epel/steering#347
No description provided.