Le mar 28/10/2003 à 23:03, seth vidal a écrit :
On Tue, 2003-10-28 at 16:58, Nicolas Mailhot wrote:
> Le mar 28/10/2003 à 22:08, seth vidal a écrit :
> > > If yum could handle a large mirror list all by itself that wouldn't be
a
> > > problem. Better fix yum than add a gui/tui bandaid.
> >
> >
> > Why can't it handle a mirror list now?
> >
> > just use multiple baseurls in each repo stanza. It will go from mirror
> > to mirror, if failovermethod=roundrobin then it select one at random,
> > first.
>
> Random is a pretty poor euristic:(
for the first time it is run, how would you want them to be selected?
Response time (will catch close mirrors at least) ?
Anyway the first time is difficult to get right - people that actually
choose mirrors do so based on their past history, so that's what yum
should do - scan all at first, collect stats, and get better choosing
mirrors over time.
at some level the user should have some involvement in what mirrors
are
better for them.
Very often, the user does not care or does not have info enough.
Whatever checks the user is supposed to do right now, a preogram can do
too (except this url is prettier/easier to type rules, but do we need
them ;)
Cheers,
--
Nicolas Mailhot