Craig,
> From a technical perspective, people whose judgement I respect
> don't think that NIO will really help a servlet container much.
> Of course, you can't prove that assertion until you actually
> implement it and benchmark it ...
One of the things that I took away from from "Mr. Sun's" presentation is
that a blocking thread-pooled solution performed quite well. A non-blocking
solution has the benefit of requiring vastly fewer threads to handle the
same number of active connections. The other benefit that NIO adds is
improved buffer handling.
I have not looked at the connector protocols, but I'd expect to find a mux
with the web server rather than opening lots of connections. Therefore, I
would not think of connection count as an issue.
That about right?
--- Noel
--
To unsubscribe, e-mail: <mailto:[EMAIL PROTECTED]>
For additional commands, e-mail: <mailto:[EMAIL PROTECTED]>
