David W Forslund wrote:
>Thomas Beale writes:
>
> > Do you mean being able to inherit across languages? I thought it was
> > wierd too, but I have seen a demo program in Eiffel...
> >
>
>But why couldn't this be done just as well with Java? Why all of
>sudden with C#?
>
I suppose no-one has developed the infrastructure to make it work. You
know Sun's ideology: Java or nothing (I'm sure everyone remembers
McNeely's absurd statement that there is no reason to program in
anything but Java.) Sun believes in a monoculture.
>
> > Two answers: the linux .net (if it exists) allows one flavour of
> > cross-platform; secondly, MSIL is designed to be compiled to binary, not
> > interpreted bytecode like Java, improving performance (presumably this
> > is MSs motive). So you have a back-end cross compiler on every platform
> > for MSIL.
>
>So it is like Corba IDL? Wht is the intermachine protocol? XML?
>
THere are tools to extract the IDL equivalent, which is I guess a bit
more like an API specification (finer grained than most people would
write IDL) from the MSIL. So if I write an Eiffel.net component, and
give it to you in MSIL, you can write a VB interface to it, based on the
interface spec extractable from the MSIL.
>> I don't know all the mechanics, but it at least works with C#, VB,
> > Eiffel and COBOL (of all things). The approach seems to be to define a
> > kind of XXX-#, where XXX is some normal language. The sharp version has
> > to conform to the C# meta-model (no MI, no-variant redefinition, no
> > genericity, etc, type system etc), and then it is a case of generating
> > MSIL from such a compiler. You might say, then you are just programming
> > in C#. But I think the point is that the XXX-# variants have the same
> > syntax as their mother langauges, and can be used in the same
> > development environment.
>But this is not interoperability, as you describe it. Seems like a step
>backwards from current technology. You start with a language and then
>build an inteface from it?
>
Actually, that's very common. Consider what could happen with say the
CEN or HL7v3 standards. They both include information models (not
service models), describing the required semantics for conformance. If
one was to build a system from them, one could start with UML models,
build software, and export the API via .net. You might say: but why not
use HDTF specs. I say, well you might (should that be "should"?), but
some people might not.
- thomas
