Michael Rumpf wrote:

>
> Hi Paul,
>
> > AltRMI is getting quite mature now.  We want to have a transport that
> utilises CORBA.  (SOAP too later).  We have an Ant task that can take some
> java interfaces and make AltRMI proxies (they are transport
> independant). It
> would be nice to have a switch called "generateIDL too" so that at
> deployment of the AltRMI using app, CORBA was an option.
> >
> > We are thinking thate there is some reusable part of 'enterprise' that
> might fit the bill. Could you give me some initial thoughts...
>
> The rmi package has not been moved into the enterprise package of
> the avalon
> project, it is still located on http://openorb.sf.net under the name
> RMIoverIIOP.
>
> >  * Do you have a Java stub compiler from IDL source?
>
> Yes it has a JavaToIdl compiler...
>
> >  * Is it an Ant task?
>
> No.

But I don't think it would be to much of a problem to make an ant
wrapper (once the static issues mentioned by Michael) have dissapeared.

> >  * Could we resuse it?
>
> It depends on what you want to do. The org.openorb.rmi.compiler.JavaToIdl
is
> somewhat broken. It is dependent on the IdlCompiler from the
> OpenORB package and as the IdlCompiler it uses static variables for
> initialization. This becomes ugly as soon as you want to start the
> compiler in process a multiple times. When creating a new instance the
> static fields are not reset and you
> have to either do that manually or create some wrapper that
> starts it out of process ( as we didi with Runtime.exec())...
> Another problem is its internal design, it does not use JavaCC for parsing
> and a as a consequence it is not as mature as it could be.
>
> However, light is at the end of the tunnel as Jesper from our
> team started a complete rewrite of the IdlCompiler which will be
> extended by JavaToIdl functionality once it has reached a certain kind
> of usable state...
>
> >  * Does it have dynamic (on the fly) capability?
>
> What exactly do you mean by that ?

I think what Pete's getting at is the ability to generate stubs on
the fly - i.e. imagine the compiler mounted as a service - provide
a class, a path, execute, stubs are ready to run ... and be able to
do that repeatedly on demand within a single VM session. I think we
run into trouble with the static state on this - but to be frank I
not really that familiar with the scope of the static problem.

Maybe its fixable relatively easily.

> Regards,
>     Michael
>
>
> PS: Steve, not that elaborated but a starting point... ;-)

So this is the price I pay for stuffing around with the build file :-)

Cheers, Steve.


--
To unsubscribe, e-mail:   <mailto:[EMAIL PROTECTED]>
For additional commands, e-mail: <mailto:[EMAIL PROTECTED]>

Reply via email to