I'm not using Craig's Correspondents framework, but I'm encapsulating the
socket interface, so it should be easy to replace the underlying layer
seemlessly. I had thought of using it, but since it is part of the standard
release, I felt it would be too early to adopt it. As soon as Squeak central
decides to adopt a different (say Craig's) socket framework, I will switch
to that. In fact, I currently recreate (fake) some of the streaming support.

I've created an "intermediate" PWS fix release, which makes it much more
stable with respect to both uploads and downloads.

Currently, the application (e.g., UploadSwiki) has to "manually" retrieve
the post body and parse it. However, I plan to incorporate a set of simple
methods that would hide the complexity and allow you to specify the rules to
apply for each field. E.g., if a field is a file, execute this block. The
block would ignore or save the file in a designated directory.

The ultimate goal is to make the PWS actions/modules/servlets/handlers very
pluggable, so that addition of functionality would only require overloading
a small number of methods or supplying a different set of "plugs". Another
goal is to keep the current PWS applications running, possibly in a
"compatibility box".

And yes (answer to another email), EmailSwikiAction is in our plans. Jochen
Rick (mailto:nadja at cc.gatech.edu) is working on restructuring Swikis and
decoupling them from HTTP/PWS.

Bolot


> -----Original Message-----
> From: Bijan Parsia [mailto:bparsia at email.unc.edu]
> Sent: Saturday, May 01, 1999 11:44 PM
> To: squeak at cs.uiuc.edu
> Subject: Re: [PWS] PWS only meant for Swiki?
>
>
> On Sat, 1 May 1999, Mark Guzdial wrote:
>
> > The PWS is definitely meant for more than just Swikis.  Bolot
> > Kerimbaev has been undertaking an extensive rewrite of the
> > innards of the PWS, and is getting some very nice performance
> > numbers for just plain uploads and downloads.
>
> Mark, is Bolot using or planning on using Craig Latta's Correspondent's
> framework? I've just started playing with it and it is *very* interesting.
> Perhaps the most obvious benefit up front is that it supports Berkeley
> socket semantics which, I understand, will eliminate the simultaneous
> access problem.
>
> Cheers,
> Bijan.
>