Re: Revisiting the Gnome Bindings

[Jamie McCracken <jamiemcc blueyonder co uk>]

On Sat, 2004-09-25 at 16:43 -0700, Sri Ramkrishna wrote:
> On Sat, 2004-09-25 at 19:28 -0400, Havoc Pennington wrote:
> 
> > Personally I'd support adding pygtk *only* - because it's a very clearly
> > different kind of language than C, appropriate in different cases. So it
> > would still be useful even if we added a new "system programming"
> > language like C++/C#/Java someday.
> 
> GNOME needs an official scripting language as well (ala visual basic)
> and Python works very well.  While I'm partial to perl, I've learned to
> appreciate Python simply because writing gui is so damn simple and
> perl's OO code looks awful and hard to understand (IMHO).  Python would
> make an excellent complimentary language.
> 

I totally agree with this one for scripting purposes only. Python is not
suitable for significant application development so we still need a high
performance RAD language for that.


> > 
> > But see how quickly followups to my mail become a language flamewar
> > about how we should really add perl, or C++, or whatever ;-) I think
> > that's the core thing that might end up blocking a dependency on
> > bindings :-/
> 
> I'd let the languages fight based on their technical merits.  It'll be
> obvious in a year or two which language people are gravitating to;
> making the choice a lot easier.  Lets not force a decision but allow
> natural selection to make the choice for us.

Wishful thinking! We currently have deadlock in the Java/Mono debate so
nothing will resolve by itself there.

What is important is that we have a RAD language which ideally has all
of the following:

1) Easy to program in and has naturally clear syntax and simple
structure. (that rules out C++)

2) Has a damn good IDE with integrated form designer so its child play
for developers to use. 

3) Gives reasonable performance and resource usage so it can be used for
creating large apps. (I would suggest kernel level garbage collection
for managed languages  - I dont want to have 10 separate threads doing
garbage collection for 10 apps I have running when one kernel process
can do it for however many apps you run and therefore is faster and more
efficient).

4) Has a damn good fully introspected native component system to make up
for our woeful lack of a COM equivalent in our platform (sorry Corba sux
big time!)

5) Is a proven RAD language. (going with the D language might be too
risky?)

I also think time is against us here and waiting another year or two is
bad - we need to have consensus ASAP so that the next major version of
Gnome (3) can take advantage of RAD tools long before Longhorn rears its
ugly head. 

jamie.

> 
> sri