GSWeb/GDL2 Status (was: Objective-C 2.0 and other new features in Leopard)

[David Ayers]

Thom Cherryhomes schrieb:
> GSWeb has been pretty much the brain-child and pet of Manuel Guesdon
> and the guys at OrangeConcept, who wrote it for their own purposes,
> and build a commercial piece of software on it (which is crap, but i
> digress.), They really don't give a damn about anyone else using it
> but them.
> 
> (it's also the same situation with our EnterpriseObjectsFramework
> Equivalent...very sad.)

Well, I'm not a guy at OrangeConcept nor is Matt Rice or David Wetzel,
and I doubt that the original author of GDL2 Mirko Viviani was an
OrangeConcept guy.

All active users of GSWeb that I know of other than OrangeConcept used
GSWeb to port WO45 Applications to GNUstep.  When we started it was
easier to port to GSWeb than to SOPE which was using a older API of EOF
(at least at that time, I'm not sure if that has changed).

I hope that neither I nor Matt give the impression that we don't give a
damn about anyone else using GDL2.  In fact many of Matt's contributions
EOInterface/DBModeller/GORM palette/SQLiteAdaptor/... are all things I
don't use at all yet I help where I can...

I also don't think that Manuel doesn't give a damn.  He has shown
significant effort to us help out on issues where we expected different
behavior than his application needed.  I do believe that the language
barrier can be very significant though.

> Yes, I am being instigatory, if for no other reason than for those
> application developers to come and speak to us.

Mission (partly) accomplished, but I doubt it will have any invigorating
result wrt code.

I believe that the main issue is lack of development resources to
allocate.  I'd be happy if someone, who could commit some time to GSWeb
could address whatever issues users may have.

Things that I'd /like/ to see are:

- s/GSW/WO/
with a GSW compatibility infrastructure similar to what I did for WO
compatibility, if this is still needed for production code)

- minimize logging:
- removal of dead code:
it's obfuscating the code for my taste, but since that's only a personal
opinion...

- drop the generic IMP caching:
this would have to be balanced with some optimizations for measurable
benchmarks.

- simplify code paths:
I feel there is excessive use internal message dispatches [yet some may
need to be retained for compatibility].

But anything we do should still work with current production
applications or suggest a reasonable new approach.  And this just
requires commitment of resources.

If you think you can take on the task of maintaining this, I'd suggest
to post on GSWHackers and talk to Dave W. who has currently done the
most work with GSWeb, but also lacks the time to needed for active
development.  But I do believe access to WO45 and some experience is a
prerequisite to insure that we don't produce yet another branch/fork.

> 
> On Nov 18, 2007 12:42 PM, David Chisnall <theraven@sucs.org> wrote:
> 

>>I'm a bit late replying to this, but is GNUstepWeb still alive?  I've
>>tried to use it a few times, including just now and found:

Alive? yes... Kicking? no.

>>- There is no tutorial at all.  The documentation (and I use the term
>>in the loosest possible sense) says 'read the Apple WO4.5 docs.'
>>Unfortunately, these all talk about using WebObject Builder, a GUI
>>tool that doesn't appear to have a GSWeb analogue (maybe there is one,
>>but nothing I could find told me about it) and so are completely
>>useless.

Indeed we could use a tutorial... I think Dave W. started with:

http://wiki.gnustep.org/index.php/Creating_A_Simple_GSWeb_Application

but got side tracked.

>>- The example code is confusing at best.  It does weird things like
>>implement classes which contain nothing but methods which do nothing
>>except call the superclass implementation.  Presumably this is done
>>for a reason, but I have no idea what it could possibly be.

Maybe you could specify... yes we some concrete subclasses (like the
class specified as WOApplication and WOSession subclasses to
instantiate) that rely on the superclasses default implementations, but
where are we implementing methods merely to call super?

Ahh Application.m in WebBookStore1 ...OK yes that was intended so that
you see the development of that method to subsequent examples that I
never got to implement.  My bad.

>>- The web page hasn't been updated in some years.  Checking again now,
>>it appears that the site now redirects to an almost content-free page
>>on the GNUstep Wiki.
>>
>>The documentation for Seaside is pretty lacking, but I could make a
>>simple web app using it quite quickly and then play at extending it.
>>With GNUstepWeb I simply have no idea where to begin.  Seaside also
>>has some really nice integration with Scriptaculous, which makes it
>>easy to produce shiny-looking web apps.  Is there any equivalent for
>>GSWeb?  I would love to be able to recycle model objects written for
>>GNUstep and just bold on a new GUI to turn a desktop app into a web
>>app, and this could be a killer app for GNUstep, but after reading
>>what passes for documentation with GNUstepWeb I still have no idea if
>>this is possible.

As for creating an GDL2 Based framework to be used with a
gui/EOInterface Application and a GSWeb simultaneously, I'd say it's
possible but not trivial.

If your looking for simple IDE to support that kind of integration in a
straight forward simple way, then, no I don't see the current developers
supplying that anytime soon.

That said, I do believe that it can be done but will probably require
another abstraction framework.  Something Renaissance-like or something
using Renaissance with main purpose of generating or parsing Renaissance
XML and creating GSWeb html/wod [potentially on the fly].  But that's
pretty much a pipe dream right now.

Cheers,
David

PS: To post to GSWHackers, I believe you may need to subscribe.
http://lists.turbocat.net/mailman/listinfo.cgi/gswhackers

PPS: [I admit I'm not fond of having this "external" list with hardly
any activity, OTOH it was created to filter the discuss- traffic as I
believe neither Dave W. nor Manuel are on discuss-.  So if I miss a
message on discuss, it most likely won't reach the relevant folks.]