Re: Scripting in Gnome

[Bill Haneman <Bill Haneman Sun COM>]

jamie wrote:

> On Thu, 2004-02-05 at 15:30, Bill Haneman wrote:
>
>  
>
> > > I was thinking along the lines of having XML as the high level
> > > definition. That could then be translated via XSLT to IDL if you needed
> > > it. I'm not saying scrap corba completely cause we do need it for
> > > backwards comatibility. If you look at the way MS is heading - its
> > > effectively using xml and web services to replace IDL in .Net
> > >
> > >      
> > That doesn't make it a good idea ;-)
> >
> > Making the XML the 'high level' definition stands the whole business on 
> > its head.  IDL was designed to express interfaces, XML is a very poor 
> > fit for this.  The fact that it would junk existing GNOME service 
> > definitions is the real stopper though, IMO.
> >    
>
> I was suggesting a future bonobo implementation that could make use of
> several back ends including the existing corba one. Therefore existing
> apps would not be broken. New apps that required corba features like
> network and platform transparancy would still get it if they needed it.
> However a new widget that used bonobo and did not require corba could
> then use an alternate (and faster) back end like D-Bus. The choice of
> back-end could be left to the app.
I don't think you can leave the back-end choice to the app; since we're 
talking about IPC and inter-application services, the clients and 
servers will (at some point) have to talk to the same back-end 
protocol.  That doesn't exclude things like bridging services of course, 
but you get my point.  Otherwise you have incompatible back ends, so the 
apps think they are speaking the same language but in reality they 
wouldn't be able to talk to one another. 

ORBit2 has a similar situation with its multiple backends for pushing 
bits around, which it deals with via IIOP: i.e. the ORB determines what 
protocols will be used to talk to peer services based on their 
capabilities, so it can be done - but not at the application level.  
Part of the reason for building a good IPC/object component model is so 
that apps don't get involved in the backend details.

> I suggest XMl as a high level wrapper for a high level interface. If you
> want to bypass the higher level and use a lower level interface like IDL
> then that should be fine too.
I do not in any way see XML as high level and IDL as low level. 
IDL is about as high-level as you can get and still accomplish anything 
useful; so any XML representation would either have to be lower-level, 
or at the same level of granularity.

This is my point about XML/IDL:   the kinds of interfaces we are 
interested in require something approximately like IDL in order to 
describe them.  Any attempt to do this in XML would in effect be writing 
an XML version of IDL, which doesn't seem particularly useful to me.

> To clarify, I am in no way suggesting anything that would break or make
> incompatible existing code.
>  
Changing the normative binding/definition level for a component or 
interface model will break existing code, if not immediately, then 
relatively soon - it would have the effect of deprecating/obsoleting all 
the existing interfaces which were based on the old definitions.

There has been plenty of discussion in the past of changing Bonobo's 
back end, but one thing that I don't see any justification for is 
changing the language in which Bonobo is defined, which is IDL.

- Bill

> jamie.
>