What's the problem of using idlj instead of writing our own...?

As anders mentioned, if we can workaround its limitation, then I would suggest 
using that... 

Is the problem that it cannot be easily extended to solve any bugs that might 
arise?

- Balaji

-----Original Message-----
From: Lars Kühne [mailto:[EMAIL PROTECTED] 
Sent: Monday, June 26, 2006 5:08 PM
To: [email protected]
Subject: Re: Specs, IDL compiler thoughts

Dain Sundstrom wrote:
> On Jun 24, 2006, at 2:43 PM, Alan D. Cabrera wrote:
>
>> Anders Hessellund Jensen wrote:
>>> Sooner or later, we need to get rid of all our generated java 
>>> classes and start generating these classes from the IDL sources.
>>>
>>> We therefore need to decide which IDL compiler to use for this.
>>>
>>> We have had some bad experiences with JacORB 
>>> (http://issues.apache.org/jira/browse/YOKO-34), so this time i 
>>> suppose we stick to the SUN-provided idlj compiler.
>>>
>>> Geronimo has a CORBA 3.0 spec project. I think it would be 
>>> reasonably easy to adapt these files to work with idlj. IMO, this 
>>> would be the best approach.
>>>
>>> One issue with idlj is that it does not have support for the "local" 
>>> keyword. I guess we will have to live with this limitation, and 
>>> treat all interfaces as non-local.
>>>
>>> What do you think?
>>>
>>> Regards,
>>> Anders
>> Can we not get JacORB to fix that bug?
>>
>> BTW, should we release the maven plugin?
>
> Can't we just write one

I like the spirit of that word "just". Nothing easier than writing an 
IDL compiler... :-)

This is already tracked in YOKO-27. Should we start one in the svn 
sandbox (i.e. outside trunk) and see how it goes? In particular, it 
needs to attract a few developers, so please speak up if you're 
interested in contributing.

> or maybe IONA could donate one?

Their current position is they won't do that, see Adi's message on 31/5 
in the "IDL to Java compiler" thread.

> Alternatively, I think openorb has one, but IIRC it needs a lot of work.

Forget it - it has a hand written parser and so many bugs that it's 
easier to write a new implementation from scratch.

Regards,
Lars


Reply via email to