On Wed, 10/15/03 at 19:31:22 +0100, Toad wrote:
> On Wed, Oct 15, 2003 at 01:26:34PM -0500, Brandon Low wrote:
> > That sounds like a pretty good idea overall, things to consider:
> > -How long does the rejection last?  Sometimes a node (mine for instance)
> > gets locked up for unknown reasons, but will be back and fully
> > functional within 5 minutes, if we start rejecting all queries from a
> > node we should make sure we allow again when it becomes contactable (or
> > remove it from the deny list once all connections to it close)... a bit
> > complex.
> 
> Initially I am assuming that the node has NEVER been successfully
> contacted, despite several attempts... we will continue trying to
> contact it. I'm not absolutely sure how this would interact with
> message queueing - if we have tried to contact it, we've wanted to send
> something to it, so presumably we accepted a request at some point.
> 
That was exactly the thought process I went through... how do we know to
keep trying to contact the node if we don't accept requests...

> > -How does this change effect overall success rates?  Sometimes a node
> > may be firewalled (accidentally or on purpose) but have worthwhile data
> > that we can eventually get from them via the connections the open to us.
> 
> There is an argument that such behaviour is malicious. It certainly is
> not conducive to path folding, which remains an important mechanism...
> However, "hostile environment" nodes would operate in exactly that
> manner, so maybe we should accept requests from nodes in our routing
> table as long as they keep adequate connections open to us.
> > 
That's not a bad idea... accept requests from the node _until_ we start
queueing and trying to open connections but failing, drop requests until
a connection becomes free or the other node opens another connection.

--Brandon
_______________________________________________
Devl mailing list
[EMAIL PROTECTED]
http://dodo.freenetproject.org/cgi-bin/mailman/listinfo/devl

Reply via email to