[Date Prev][Date Next]
At 09:10 AM 6/5/00 +0200, Julio Sánchez Fernández wrote:
>> I haven't looked at how easy it would be to push the thread-pooling
>> code "outside" of the platform-specific thread code. Any comments
>> here? Preferences?
>Only a comment if you are into threading issues. The main thread may
>be pretty busy when cryptography is used. Maybe you want to look into
Not to meantion other bottle necks in the listener thread.
I'd really like to move to a system where the listener thread
was not responsible for the actual input/output. The listener
would just manage the event loop and assign a worker thread
to any socket needing attention.
However, this requires a significant redesign which is likely
best deferred for a bit. I rather Mark just implement some simple
pooling (at an existing thread_create() boundary) as a first step.