On Sun, Dec 30, 2001 at 11:58:53PM -0600, Edgar Friendly wrote:
<> 
> I agree that the idle timeout should be decreased.  I disagree with
> your analysis of the 0.3 network.  The 0.3 network worked because
> there was no pre-requisite for contacting a node (like knowing its key
> is in 0.4), so you could contact nodes willy-nilly and you'd always
> connect.  Now there's the problem of nodes changing identities and a
> host of other connection errors that can occur. 

Not true at all - 0.3 nodes did not find others "willy-nilly" but by the
same procedures that 0.4 nodes do. The only difference between a 0.4
reference and an IP is the bytelength - some nodes may change
identities and thus become uncontactable, but nodes may also change IP
addresses or whatever to the same end. 

The big difference is the complexity of the session management. In 0.3
we used a simple request/response management for connections - while any
message could in theory go down any connection, the nodes always
established new ones for requests and generally replies would come back
through that same connection. The way this is done in 0.4, with the
connections being completely independant to the messages sent over them,
without even a temporary role as server or client, is really quite a
radical protocol design - I'm not aware of any other protocol in use
that does it - and it has proven a lot more complicated to decide when
to keep connections open and when to close them then was anticipated.

There are also external problems effecting the connection management -
such as the chronic problems with the datastore and the large amount of
messages generated by the infinite restarts for example.

<>
-- 

Oskar Sandberg
oskar at freenetproject.org

_______________________________________________
Devl mailing list
Devl at freenetproject.org
http://lists.freenetproject.org/mailman/listinfo/devl

Reply via email to