Re: [apache-plusplus] Re: Process model for C++ Apache

[Bret <mersault@busboy.sped.ukans.edu>]

>> Michael, I agree, we're approaching novel length here.
>
>We seem to be alone in this discussion and I think we've
>flogged this novel enough. Since no one else is interested,
>let's give the novel a graceful burial. I gratefully
>appreciate your comments, as well as Dean's comments -
>They highlighted several oversights in the design of
>my project. I certainly did not give proper consideration
>to the performance effects of the different threading
>models.

	I'll offer my thanks as well... I've learned a great deal from everybody's
contribution, namely that I have a hell of a lot to learn about threads... :)

>This discussion probably should have occurred in new-httpd,
>since most of the discussion was general, and very little was
>"plusplus".
>
>Also the process-model discussion should have
>been preceded by an architecture/product discussion - its
>hard to consider design issues such as threading models
>or class definitions without having a clear idea of how Apache
>fits or should fit in the larger picture of distributed
>computing. I believe the role of http servers is/will be
>very different now (or in the very near future) than how
>they are currently conceived. No reason Apache shouldn't
>take the lead in these future roles as it has the lead in
>the current conception. I take the blame for starting the
>process-model discussion before the architecture/product
>discussion.

	Actually, I think that can be evenly distributed.  I caught myself wishing
several times that we could take our discussion and apply it specifically
to a C++ implementation of Apache.  So...
	While I agree that our thread on threadings should die a graceful death, I
would like to return to the notion of a C++ implementation of everybody's
favorite Web server.  Michael raises some valid criticisms... we need
further discussion of architecture and general design before deciding on a
thread package (I'm still favoring ACE... just kidding, Dean. :) ).
	With that in mind, where to begin?  Should we focus on Dean's NSPR
implementation, improve NSPR to determine thread models at run-time, throw
the whole thing into objects and go?  Or should we study JAWS a bit more
carefully and build something up from ACE?  Either way, we really should
begin discussing the various issues involved...
	Comments are not only welcome, but encouraged...

									- Bret -