Comment 3 for bug 368312

On Wed, Apr 29, 2009 at 12:32 PM, Matt Nordhoff
<email address hidden> wrote:
> FWIW:
>
> I'm not sure I agree. If my bzr server is down for half an hour for
> upgrades, I don't want it to look like all of my branches up and
> vanished. Plus, I occasionally look at my branch listing page to make
> sure none are broken.
>

So, we could possibly have a time delay, so that branches that have
only just broken are still on the main listing.

As for the latter, well, this will actually make it more obvious that
no branches are broken -- there'll be a link that shows the number of
broken branches, and probably a stable URL for the broken branch
control page.

>
> A page for broken branches sounds useful, though. Especially if it
> provides a button to remirror them all without having to click on every
> single branch! (I think I filed a bug about adding such a button?) I
> hope it'll handle having dozens of broken branches as well as the normal
> listing page does.
>

Yeah, mass operations are pretty important. It might not have them in
the first pass, but it's definitely something I'm conscious of.

> Wait, do you want to remove branches that've just had one failed mirror,
> or branches that have been failing long enough that the puller has given
> up on them? The latter would be less problematic, but I think it's still
> happened to me a couple times when something goes wrong in bzr+http.
>

I hadn't thought much about it. Probably ones that the puller has
given up on, plus any hosted or imported branch that is broken.

I guess better notification of mirroring failures would address some
of your concerns.

Thanks for the feedback!
jml