check fastest mirror

[Jean Michel Bruenn]

Hello Striker,

the main idea behind it, is still good. Other distributions are doing it the same
way. Traceroutes gives the shortest way to a server - I mean, if i have only 3 hops
to one server it's nearer then one Server with 25 hops (I know, it could be possible
that the server with 25 hops is faster) but we could give the user a chance to choose.

I mean for example - If we have such a script it could print out the 5 nearest servers.
The User could choose one of them - It's not needed to use this automatic script.

And at least, ppl who wants, can choose their mirror manually. I tested this script and
it was right (if we don't look at this blocked tracerroute things) i got at least
faster mirrors - try it, use the mirrors the script will print out for you.

And, one question, if this is so bad, why are other distributions using it? (Yeah, all
other distributions are bad... hello?)

There aren't many ways to do what we want, we could simply add for each mirror a prefix
like DE, EN, US (maybe N_US and S_US for North and South) and things like that, and at
installation and if the user like in a lunar prompt, let him choose from which country he
is, and then display the mirrors so he can choose what he likes to have.

like cpan it does - If u install cpan and it asks you for manually/automatically
installation it will ask you for a mirror list - first you give in which continent
then which country, and then the mirror itself.

We could do it like this, too, and optionally print out what's the nearest server.
You're right, that does not mean it's the fastest server. But i think in 90% it will
be the fastest one.

Cheers
Jean


On Mon, 28 Aug 2006 09:58:21 -0500
Jon South <striker at lunar-linux.org> wrote:

> Jean Michel Bruenn wrote:
> > This script uses netselect to determine the fastest available mirror. The main problem is, if a mirror is blocking traceroutes it's shown as slowest server, so it's possible that this tool writes that the fastest server is in the netherlands (if i live in germany and all german mirrors blocking traceroutes). On the other side, that's good - Cause it would use only working mirrors.
> 
> Your main problem is not blocked traceroutes, your problem is using 
> traceroutes as your guage for the tests. Traceroute, pings, etc, are all 
> low priority packets to routers, so using them as a basis for testing 
> what is "the fastest" is just a terrible idea in the first place. This 
> concept was discussed the last time it was brought up and it still has 
> the same problem as before: it's just unreliable.
> 
> -Striker
> _______________________________________________
> Lunar-dev mailing list
> Lunar-dev at lunar-linux.org
> http://foo-projects.org/mailman/listinfo/lunar-dev