David Walser <[EMAIL PROTECTED]> writes: > fpons - any comment on this? > > David Walser wrote: > > --- David Walser <[EMAIL PROTECTED]> wrote: > >> --- Guillaume Cottenceau <[EMAIL PROTECTED]> > >> wrote: > >> > David Walser <[EMAIL PROTECTED]> writes: > >> > > I'm writing, because if the location [rpmdrake is] > >> downloading > >> > from > >> > > is the same as the proxy, it should not use the > >> > proxy, > >> > > as that sometimes plain doesn't work. > >> > > >> > that's curl's job I suppose.. > >> > >> No. Plus you'd have to do it upstream in wget too. > >> Anyway, Unix tools in general let you do stupid > >> things > >> if you really want, including downloading from a > >> host > >> through a proxy on that host, even if that doesn't > >> work. With rpmdrake though, it's kind of doing a > >> stupid thing for you w/out you explicitly asking it > >> to, as it's picking up that http_proxy variable. > >> rpmdrake should check for this. > > > > Ok, we're both wrong. It's not curl or rpmdrake's job > > to check for this, it's urpmi's. > > > > Imagine this scenario, you've got a proxy configured, > > which is also the location of one of your urpmi > > sources, but you have sources from other servers too. > > You run a urpmi command to install some packages, some > > which will come from each place. If you don't do > > http_proxy="" first, the ones from that server won't > > work, but if you do, the ones from the other servers > > won't use the proxy. > > > > So if urpmi is installing something from the same host > > as the proxy, it shouldn't use the proxy for that package.
Proxy by media is the best... yes I agree. François.