To: devl at freenetproject.org
Subject: Re: [freenet-dev] (no subject)
References: <20030430023525.A20EC181BA6 at thelema.dyndns.org>
<20030502141713.GB3456 at amphibian.dyndns.org>
From: Edgar Friendly <[email protected]>
Date: 02 May 2003 15:28:49 -0500
In-Reply-To: <20030502141713.GB3456 at amphibian.dyndns.org>
Message-ID: <m2fznxf5z2.fsf at thelema.dyndns.org>
Lines: 31
User-Agent: Gnus/5.09 (Gnus v5.9.0) Emacs/21.3
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Toad <toad at amphibian.dyndns.org> writes:
> > My suggestion (and I hope people realize this) assumes some framework
> > behind the scenes to actually have connections to route over.
> > Optimally fred would keep active one connection (or two, if we can't
> > multiplex
> > over a connection) to each node in the routing table. Suboptimally,
> > some limit on # of connections and a method of choosing what
> > connections to attempt so that normal routing is subverted the least.
>
> Well, because of our threading model, we are limited to 48 connections.
A lot can be done with 48 connections. But if you insist that's not
enough, then I guess I have to add another assumption: a threading
model that can handle a good number of connections. (BitTorrent
scales to almost 1000 connections per client)
> And we open a connection when we route to a node that does not have a
> connection open. How do you propose to decide which nodes to open
> connections to, without actually opening connections to them? :)
>
I propose the node keep counts of which nodes we would want to route
to, but aren't because there's no connection. Once there's enough
'missed' opportunities, drop the least used connection to make room
for this one.
Thelema
--
E-mail: thelema at swbell.net Raabu and Piisu
GPG 1024D/36352AAB fpr:756D F615 B4F3 BFFC 02C7 84B7 D8D7 6ECE 3635 2AAB
_______________________________________________
devl mailing list
devl at freenetproject.org
http://hawk.freenetproject.org:8080/cgi-bin/mailman/listinfo/devl
