Re: [Mldonkey-users] Re: can t connect to server

[Pierre Etchemaite]

Le Sun, 24 Nov 2002 23:15:11 +0100, Roland Arendes <address@hidden> a
écrit :

> We need the source_propagation fixed. It seems that this is the issue
> of the filled up upstream. I read that sources received that way are
> immediately tried and not queued. Perhaps thats the bug.

Immediately trying sources is not optimal, and there's a patch that fixes
that already:

http://savannah.nongnu.org/patch/?func=detailpatch&patch_id=700&group_id=1409

But that doesn't fix the "fill upstream" problem. I think that would require
fixing 2 or 3 things in the rate limiter, and give a separate limit (5% of
traffic ? more ?) to UDP alone, for example.

> And while we're fixing this, it could be good for the network to
> make it compatible with the emule source propagation.

Since I guess eMule protocol extensions will be as documented as mldonkey
ones, can someone read the eMule sources and write down specs ?
Can one implement some of eMule features without having to implement all the
others ?

BTW, the whole eDonkey network is in bad need of written down specs,
validated by all participants (client writers, server writers, indexing
service writers...). I they were specs, we wouldn't wake up in the morning
and find that our beloved client cannot connect anymore.

> Anyone knows when mldonkey will come back? We need development. :-)

If you want something more "bleeding edge" than CVS 2.00+2, you can try
this patch:

        http://concept.free.free.fr/mldonkey/

But some fixes may require mldonkey, at least if you want them soon ;)

> M> I guess the problem is the server list (over 2k servers...these time
> M> are over my friend...)
> 
> Well, the server list of mldonkey was every time I can think of way
> greater than it should be. I can remember times of 25k+ servers (22k
> of them were dead, though).

Already answered in this list, list cleaning is not done when less than 200
servers would be kept, so yes that another problem caused by Ludgunum.