Re: [DotGNU]Choice of bytecode spec (was: Support Java for .GNU)

[David Sugar]

Personally, I think we need to do both.  We should make it possible to use the 
free Java stuff (KAWA being my immediate example) as a runtime bytecode 
platform by adding appropriate Java DotGNU class libraries, in addition to a 
fully cleanroom implimentation of the ECMA spec byte code system and C# 
language that Microsoft has submitted for standardization.  There are 
different reasons for each; Java offers a mature platform already in place 
that offers alternatives not controlled by Microsoft, and having a fully ECMA 
spec compliant platform is also important should it become mandatory or 
reguired in commercial and/or govt bids for software, as well as a tool that 
can be used to drag Microsoft into continuing standards compliance.

The idea of over time extending KAWA's or similar Java virtual machine systems 
to support CLR had actually been considered and reviewed by some of the KAWA 
developers, and for various reasons, this approach, (that Hanson seems to 
suggest), was not felt to be feasible by them.  Of course they could be wrong, 
but I think the two (CLR and JVM) should in any case be seperate bytecode 
platforms usable within DotGNU, especially in case on of these approaches 
founder because of the legal issues involved (Microsoft, in case it plays 
dirty tricks and/or withdraws from ECMA entirely, or Java, because Sun is so 
hard to deal with in general...)

David

Norbert Bollow wrote:

> I wrote:
>
>
> > >   this sounds very much like anti-Java FUD to me.  Please be
> > > specific.  What are the problems of the plan that has been
> > > proposed?  How would it "pollute our system with legal baggage"?
> > > (Nota bene, the proposal is to start with a free JVM
> > > implementation)
>
> Martin Coxall <address@hidden> replied:
>
>
> > My career is Java. I have no need to FUD.
>
>
> Then please provide the requested specifics.
>
>
> > What I am worried about is blindly allowing an implementation
> > of a privately owned set of specifications forming the core of
> > a very important free software project.
>
>
> I can assure you that we're not planning to go about this
> blindly.  We know that the choice of bytecode is a crucial
> decision for the project, and the time has come to look at
> the various reasonable options.  My current list of reasonable
> options is this:
>
> a) Johan Hanson's proposal
>     
>      Start with Java bytecode, the Java language and the Java libraries.
>
>      Then, extend the bytecode and runtime environment to have
>      features that are supported by Microsoft's bytecode.
>      Design the modifications to the bytecode so that a recompiler from
>      CLR bytecode to DotGNU bytecode would be as simple as possible.
>
>      Start with Kaffee or some other existing Java implementation that
>      is licensed under the GPL. (if any)
>
> b) Use and extend Portable.NET
>
>      Endorse IL and C#, extending them as necessary.  Use and
>      contribute to Rhys Weatherley's efforts to write a good
>      C# compiler and runtime system in C
>
>      The website of Rhys Weatherley's Portable.NET project is
>      at http://www.southern-storm.com.au/portable_net.html
>
> I would like to invite everyone to make suggestions for items
> that maybe should be added to this list of "reasonable options".
>
> Greetings, Norbert.