Is it that the applications were merged, or that audit is just better exposed via a textbox on Diffusion now? I still see Audit as an independent application (currently uninstalled).

If it's still its own thing, perhaps this textbox on Diffusion should only show up when Audit's enabled? That would be the most reasonable course of action to me.

Thanks to T77959 I have learned that trying to comment in a commit results in 404, at least in rEADT7b77eb7970283304554888e2bcf88bcda1f57f0e.

That might be a good interim solution to avoid people commenting in places that might disappear soon.

In T75715#836328, @Qgil wrote:

Strangely enough, the 404 makes this task Normal priority.

The 404 is away, and I was just able to post a comment in rEVEDa2338df67535c7909c045b00fc047021df532fc5

I don't think this is good...

Considering Audit and Diffusion are slowly becoming the same application I'm not too worried. I'm also not sure what the harm is in allowing people to audit commits; it seems like a useful feature to be able to flag a commit for review after it's already been merged.

If we think that opening this new channel causes no harm, then we can resolve this task as Declined, and document the feature so people know what to expect.

The feature in itself is ok, but I agree with Qgil that having yet another place where to comment a commit is disruptive: we don't want to further fragment conversation.

Additionally, if we ever wanted to import CodeReview discussions into diffusion for commits imported from SVN, having recent discussions to merge to the old ones seems likely to complicate things.

Isn't there an option to restrict write permission to admins/nobody?

In T75715#993013, @Nemo_bis wrote:

Isn't there an option to restrict write permission to admins/nobody?

I think there was a bug here we'd identified. @chasemp?

Without https://secure.phabricator.com/T6630#85907 I am not sure what we can do,. We can tinker with old audit perms to see how they affect.

Personally I find that audit is highly useful for release engineering: When some fatal has slipped through the cracks and I discover an error in the fatalmonitor, I can track it back to the responsible code and raise a concern to get the problem dealt with. See for example: rECOE36abbc628834ae141f9c28f0bc3d2c8d44524754

In T75715#1423558, @mmodell wrote:

Personally I find that audit is highly useful for release engineering: When some fatal has slipped through the cracks and I discover an error in the fatalmonitor, I can track it back to the responsible code and raise a concern to get the problem dealt with. See for example: rECOE36abbc628834ae141f9c28f0bc3d2c8d44524754

Except the comments you left there and any events surrounding it were never seen by anyone because I'm not looking there. I didn't see your comment pointing out the problem until two days after I fixed it myself independently.

In T75715#1423592, @Krinkle wrote:

In T75715#1423558, @mmodell wrote:

Personally I find that audit is highly useful for release engineering: When some fatal has slipped through the cracks and I discover an error in the fatalmonitor, I can track it back to the responsible code and raise a concern to get the problem dealt with. See for example: rECOE36abbc628834ae141f9c28f0bc3d2c8d44524754

Except the comments you left there and any events surrounding it were never seen by anyone because I'm not looking there. I didn't see your comment pointing out the problem until two days after I fixed it myself independently.

Are you sure you aren't talking about a different commit? rECOE36abbc was resolved just a bit more than an hour after I raised a concern.

I can't think of any situation where I raised a concern and then it went un-addressed for any significant amount of time. When I'm deploying code and a problem comes up, I don't just raise concern and then wash my hands of it, I investigate the problem, contact the appropriate people on IRC, etc. If something is serious enough that I raise a concern then letting it go for two days would be a fairly serious failure on my part. Is there a different commit that I'm not seeing?

I'd actually propose declining this outright.

+1 for decline.