Re: Scripting in Gnome

[Geert-Jan Van den Bogaerde <geertjan gnome org>]

On Thu, 2004-02-05 at 20:40, jamie wrote:
> On Thu, 2004-02-05 at 19:31, Geert-Jan Van den Bogaerde wrote:
> > On Thu, 2004-02-05 at 20:16, jamie wrote:
> > > On Thu, 2004-02-05 at 18:45, Sean Middleditch wrote:
> > > > Why would XML work better than IDL?  IDL could just be compiled to
> > > stubs
> > > > that use D-BUS.  The "oh, it's XML, so it's just going to be easier
> > > and
> > > > better" rehashing is getting old.  Especially given detailed claims to
> > > > the contrary by others in this thread.
> > > 
> > > Whether you use XSLT or a built in xml parser, its quicker to develop a
> > > programme to read in and process xml than plain text - okay? Im fed up
> > > repeating this but seeing as you are exceptionally abrasive I thought I
> > > better say it once more in the hope of getting my message across. XML
> > > can also be validated against a schema/dtd and thats handy too for
> > > debugging whether your produced IDL is correct etc.
> > 
> > This argument might have carried some weight if the limiting factor in
> > producing language bindings was the effort of writing the IDL parser for
> > that language. It is not, because parsing IDL is only a very small part
> > of what is involved in writing language bindings, so it's not worth
> > deviating from the established standard of CORBA IDL over.
> 
> 1) You can automate and verify the conversion using XSLT - its gotta be
> a lot quicker than writing a parser, performing error checking and doing
> the conversion.

[Note: My last e-mail quoted above did not get posted to the list
because I wasn't subscribed to the list under this address. It's quoted
above in full so I won't repost it.]

If you're talking about generating the stubs from your XML-IDL through
XSLT, then as others in this thread have mentioned, XSLT quickly gets
very ugly when doing XML -> non-XML conversions. It's just not very well
suited to this task.

> 2) I was only suggesting deviating if bonobo no longer became dependant
> on corba and therefore there was no longer a strong case for using IDL.

OK, perhaps it would be useful if you prepared a more detailed
technological description of the system you propose. I think a lot of
the criticism you're getting is coming from the fact that you're not
explaining very clearly what benefits your solution would bring, apart
from  vague unsubstantiated claims of "faster" and "more consistent".

Geert-Jan

> 
> jamie.
>