Re: [Mldonkey-users] search requests blocked

[Fortin Denis]

Le mar 30/09/2003 à 14:47, Roland Arendes a écrit :
> Hi
> 
> > That's interesting. I also get this "result" *in addition* to my real
> > results.
> > If you can identify the servers that block you search requests you can
> > probably
> > blacklist them (not sure you can, but you get the idea) and use other
> > servers
> > instead.
> 
> You shouldn't do this. I suppose these messages are sent by the complete
> new and even better eserver (successor of dserver). Lugdunum wrote a new
> hell of a server from scratch (16.40). If you blacklist these servers,
> you're quite stupid (Razorback 254k Clients, Probenprinz 200k Clients,
> ..).
> 
> I think someone should hotfix the situation and release a 2.5-4 next to
> the devel version. Furthermore I think this message is caused by this:
> 
> - keepalive pings are most of the time piggybacked into other services
> messages: lower TCP bandwidth (using less IP packets)
> - Extension of the UDP protocol to send many Search results per packet.
> - Extension of the UDP protocol to allow clients to ask sources for many
> MD4 hashes at once, instead of one UDP packet for each MD4. (should reduce
> UDP traffic by a 2 or 3 factor)
> - Tracking of too old emule versions.
> 
> You can read more here: http://lugdunum2k.free.fr/kiten.html
> 
> Old clients do not, but are still tolerated (thats why you still receive
> results).
> Changing the emule version within mldonkey would bypass and not fix the
> problem. Someone should add these capable udp messages to MLdonkey and the
> problem is gone.
> 
> (I'm a fan of Lugdunums work for the community)
> 
> Bye
> Roland

Where are the specs of these new messages?
Should we have to search in emule sources to find them?

IMHO it doesn't send this serch result to clients which don't implement 
these new packets but to really old versions of emule.
So we must discover how (and why) they mark a client as too old (i bet that 
it's when 
they fake to be a client).