[Twisted-Python] IOCP Reactor (Was: win32 buildbot)

Justin Johnson justinjohnson at fastmail.fm
Tue Mar 16 13:37:12 MST 2004


I would love to see a more stable version of the windows reactor.  :-)

Occasionally where the win32eventreactor starts spinning wildly out of
control when I do too many things at once.  I assume this is the 63
kernel events limitation or some such thing...

-Justin

On Sun, 14 Mar 2004 19:36:10 +0000, "Paul Moore" <pf_moore at yahoo.co.uk>
said:
> Itamar Shtull-Trauring <itamar at itamarst.org> writes:
> 
> > Fixes to win32event reactor or a finished IOCP reactor so we can
> > abandon it are welcome ;)
> 
> I was interested to see this, as I've been aware of the limitations of
> the standard select reactor on Win32 (no process stuff, for example),
> but given the comments in the documentation for the win32event
> reactor, I wasn't sure that was the way to go either.
> 
> One neat feature in later versions of Win32 (Windows 2000+, I believe)
> is a kernel-level thread pool, which handles async I/O very cleanly,
> as well as a lot of other stuff, such as waiting on more than 63
> kernel events, asynchronous function calls, etc, etc. Looking at the
> IOCP reactor, it doesn't seem to use this.
> 
> Some questions:
> 
> 1. Would a reactor based on Win32 thread pools be worth having?
> 2. Is lack of NT support a problem (Win9x is already disallowed by
>    using IO completion ports)?
> 3. Does anyone want to pick this up (the author of the IOCP reactor,
>    or the existing win32event reactor, for example)?
> 4. If no-one else wants to pick it up, where would I find pointers on
>    writing a reactor? Assuming there isn't a useful "what is a reactor
>    supposed to do" document (I couldn't find one) what code would be a
>    good starting point? [Note: If this is what ends up happening,
>    expect completion in something like 2038, given my copious
>    quantities of free time, and the likelihood of me managing to stick
>    with a substantial programming project :-(]
> 
> Paul.
> -- 
> This signature intentionally left blank
> 
> 
> _______________________________________________
> Twisted-Python mailing list
> Twisted-Python at twistedmatrix.com
> http://twistedmatrix.com/cgi-bin/mailman/listinfo/twisted-python




More information about the Twisted-Python mailing list