Re: shutting down httpredir.debian.org?
Hi,
On 12 April 2016 at 09:04, Peter Palfrader <weasel@debian.org> wrote:
> Hi,
>
> we keep getting reports of httpredir.debian.org not working correctly,
> such as intermittently just sending errors or redirecting to mirrors
> that are out of date.
[...]
> So, it appears as if currently nobody has time or the energy to take
> care of httpredir.debian.org properly.
>
> I suggest we shut down the service for now. If, at some future point,
> somebody wants to maintain again we can always start it up again.
Let me summarise the current situation:
- user-visible errors happen from time to time, possibly increasing over time
- the current code base has a high cost of maintenance
- the current infrastructure is not homogeneous, there are three
hosts, each launching the application in a different way
- the main code contributors (Simon and yours truly) have been
EBUSY/ENOTIME for a while - Simon, please correct me if I'm wrong
- the service has not been fully integrated into the mirror's team
workflow - ex. a recent change to ftpsync makes httpredir believe
mirrors do not handle some files correctly
- the service has not been fully integrated into the mirror's team
umbrella, notably contact addresses as weasel correctly noted
- the service has been widely welcome by users - a single address
- the service has superseded similar services: geomirror.d.n, cdn.d.n
- no other service exists that could supersede httpredir while
retaining control over it
- no other read-to-use solution exists that could supersede httpredir
without a significant cost (time)
(all points are TTBOMK and IMHO)
Given the above, I believe it to be in the best interest of the
project to further attempt to improve the service rather than shutting
it down.
What I propose is:
1. to fix the recent regressions and perform some maintenance on the
service in an attempt[1] to reduce the user-visible errors. Before the
end of April.
2. define the next steps towards improving the service - can be done
in an open way in the mirrors ML. Starting today, with no ending date.
3. organise a sprint in order to ensure that time is available to work
on the service. Perhaps during SunCamp, as people such as weasel also
appear to be interested in participating - so let's say end of May.
[1] as of this time there's no metric for this
Do people agree to the above?
PS. Regarding mirrorbrain, I can go into further details as to why it
doesn't fit our use case but I think it's orthogonal to the discussion
here.
It's a great project, Peter Pöml has done a great work, and httpredir
is inspired on it.
Cheers,
--
Raphael Geissert - Debian Developer
www.debian.org - get.debian.net
Reply to: