[standards-jig] JEP 65 - Bytestreams

Justin Karneges justin-jdev at affinix.com
Fri Dec 20 01:03:22 UTC 2002


On Thursday 19 December 2002 04:24 pm, Ben Schumacher wrote:
> Richard Dobson wrote:
> | Also to the people who say you have to wait for a TCP timeout before
> | trying the next step (typically 60 seconds) why not implement a manual
> | timer of say 10-20 seconds, if a connection is not established after
> | that kind of a period of time IMO it probably wont, most connections
> | should happen pretty quickly.
>
> I believe Dizzy already suggest this, but I agree completely. There is
> no reason that a connection that is going to work should take longer
> than 10-20 seconds. And if it would work after that longer period, then
> its likely that there are dropped packets, or other problems which would
> ultimately lead to a useless connection anyhow.

I don't think cutting the connection short is a good idea.  You might lose a 
possible connection due to a temporary network problem.  Let's consider the 
reverse scenario:  what if you had a DTCP implementation where the initiator 
would wait 10-20 seconds before trying to connect?  This is more ideal, as 
you wouldn't cut any connections short, and you spend the same in bandwidth.

But these are all implementation specifics.

-Justin



More information about the Standards mailing list