Kore developement

[Glynn Clements]

Artur Biesiadowski wrote:

> I'm thinking about contibuting to kore and I have some questions/notes
> 
> - How exactly clean-room implementation state is forced if sources for
> sun's classes.zip is available - is this just by not looking at sun's
> work at the moment of writing ?

By not looking at Sun's code.

> I understand that most classes can be changed by implementing
> slightly different variables and mechanisms, but some of the sun's
> code is most obvious one - do we need to make all things different
> or only some of them ?

Some functions will inevitably be the same as Sun's implementation;
this can't be avoided unless you look at Sun's code in order to make
them different.
 
> - I understand that static holders for native methods are temporary, but
> they cannot be done in such way - most native functions need to be private
> for security reasons.

It would be nice if the kaffe.system package could, as a special case,
be made accessible only to classes without a class loader. This would
enable java.net.* to use methods from kaffe.system.NativeIO, for
example.

Otherwise the kaffe.net.Native* classes will need to be made
non-public members of their corresponding packages.
 
> - Will kore be aimed separately for 1.0.2 and 1.1 or it will be just
> upgraded to 1.1 when time allows ?

To date, Kore has been aimed at JDK-1.0, as that is what all of the
documentation relates to.

> - As about incorporating kore into kaffe it is not very big problem.
> We need some define switch in Makefile and create kaffe_native.so for sun
> classes or for kore classes depending on it, or make some command line
> switch to choose which library will be loaded. Most of the native code
> implementation can be just copy-pasted as there are no legal problems :)

Er, well I would imagine that all of the files generated by running
kaffeh against Sun's classes.zip belong to Sun.

> Of course names of fields have to be changed, but I think that 90% of that
> should be straightforward.

I'm not sure that changing field names is sufficient. I imagine that
the structs that would be produced by running kaffeh against Kore's
classes.zip would be substantially different to the existing ones.

> It is unfortunate that my exams starts in few days, so I won't be able to
> actively help for about month, but after that I'm willing to contribute.
> 
> Questions is if you want some help, if yes, in what form ?

All contributions gratefully accepted.
 
> Is there any group making kore with some maillist or it just your work ?

Some of the work, particularly in java.lang.*, was done by George
David Morrison. He had to drop out, due to college work I believe, so
it's just me right now.

> Are you going to develop this project further ?

Depends on whether it's going to get used. I don't know anything about 
Kaffe's internals, so I'm relying on others being willing to make the
necessary changes.

I'm not willing to start cloning Sun's internal API, unless someone
has paperwork from Sun saying that it's OK to do so.

-- 
Glynn Clements <glynn at sensei.co.uk>