[arch-general] Mirrors out of date

[Simon Hanna]

Hi,

I don't think hashes are a good way to do that. The Repositories are quite
large I'm not sure what value is the right one from this [1] statistics
page. Either way, after every update the mirror would effectively block for
the time the hash is computed, since everything has to be read.

Anyway, the time a mirror isn't in a sane state should be quite short if
ever. I don't know how updates are handled, so I can't say what happens and
what doesn't. Nontheless the servers normally have enough bandwidth to sync
in a few seconds (I guess)...

The most important part of my answer is, that this mailing list isn't the
right place for discussing this topic. There is a mirror-list mailing list
out there [2]

cheers,

Simon
[1]: https://www.archlinux.de/?page=RepositoryStatistics
[2]: https://mailman.archlinux.org/mailman/listinfo/arch-mirrors


2014/1/15 Mark Lee <mark at markelee.com>

> On Wed, 2014-01-15 at 22:54 +0100, Guus Snijders wrote:
> > Op 15 jan. 2014 19:17 schreef "Mark Lee" <mark at markelee.com> het
> volgende:
> > >
> > [...]
> > > > > There is a file called lastsync. It is read <
> > > > > http://mirror.csclub.uwaterloo.ca/archlinux/lastsync>
> > > >
> > > > Yes, I see that there's a check script but what is it? I'd like to
> know
> > > > how that link calculates percent synchronization for a particular
> > > > mirror. I ask this because there have been instances when a mirror is
> > > > claimed to be 100% synchronized but isn't.
> > > >
> >
> > >
> > > Oh I see it now, so "completion" is only a percentage of the number of
> > > times the checkscript is successfully run on a particular mirror
> without
> > > disconnecting.
> > >
> > > Is the lastsync value a hash value or temporal?
> >
> > As Thomas' message explains, it's a timestamp. A very nice solution if
> you
> > ask me.
> >
> > mvg, Guus
>
> Salutations,
>
> I see. My concern is regarding mirrors that aren't fully up to date
> (I've encountered the situation before and it wasn't a packaging error
> at the time) and reporting this fact to the user. Instead of using a
> time stamp, why not use a cryptohash of a list of the files installed as
> the lastsync value. Pacman -Syy could then check the hash value (it
> grabs from the tier 1 mirror) and check it against the current mirrors
> it's using.
>
> Regards,
> Mark
> --
> Mark Lee <mark at markelee.com>
>
>