http://issues.apache.org/SpamAssassin/show_bug.cgi?id=4861





------- Additional Comments From [EMAIL PROTECTED]  2006-04-19 20:44 -------
(In reply to comment #10)
> the only thing I'm worried about, is increasing that timeout to 10 seconds. 
> given that we currently hang waiting for the response, and then razor, pyzor 
> and
> SPF are all waited for on top of that, that's a *long* time :(
> 
> what about 8 secs?

Ditto.  I was about to suggest 7 seconds myself.

If the typical response arrives in 3 seconds, I'd say (without any data to
suggest a better value) twice as long should be sufficient.  I don't think
there's much point in waiting longer than whatever ~90% of responses take.

Also, FWIW, I'd speculate (without knowing anything about how the DCC backend is
implemented) that queries more likely to get a positive response would receive a
response faster than queries least likely to get a positive response since
they'd be more likely to have the required data in memory.  I know that this
appears to be the case with April's Outbound Index service.


(In reply to comment #9)

> or when a client has sent more than 400,000 requests per day and the server 
> DoS
> defenses are delaying responses by 4 seconds are many seconds needed."

We shouldn't care about delays/timeouts caused by this.  In fact, DCC would be
better off if these queries timed out since the user might clue in to the
problem and arrange peering.



------- You are receiving this mail because: -------
You are the assignee for the bug, or are watching the assignee.

Reply via email to