Re: glib / CORBA integration examples ...

[Michael Meeks <michael ximian com>]

Hi Dan,

On Tue, 2002-09-17 at 16:34, Dan Kegel wrote:
> I can tell you my reason: conversion. 

	Yes conversion sucks; it'd be great to have more unified data
structures, particularly mapping a sequence as a GAarray, but sadly it's
not possible until an API/ABI breakage point.

> To use Corba properly, one must convert between one's own data
> structures and Corba's on every

	Yes, and that's painful; I mean to add some methods on the sequence to
make them easy to extend in the ORBit for 2.5 - not at all difficult to
do, but would make life lots easier for building sequences - if you want
to do that work, mail me.

	Ultimately if you store the stuff as a CORBA type internally, you don't
have as many problems - and it's type-safe, and frequently more
efficient than the GAlternative.

> Sure, it's not portable at the api level, but it would be at the
> protocol level.

	Indeed; one of the things we use a fair bit is the fast in-proc
bindings, for which we have a public ABI - that is based on the CORBA
sequence layout - so ... it makes it difficult to change really;
certainly we'd need to map glib -> CORBA on some in-proc calls to meet
the ABI requirements :-) which would be a pain.

	Regards,

		Michael.

-- 
 mmeeks gnu org  <><, Pseudo Engineer, itinerant idiot