From: Alexandre Oliva (oliva@dcc.unicamp.br)
Date: Thu Feb 11 1999 - 09:21:15 EST
On Feb 10, 1999, Tim Wilkinson <tim@transvirtual.com> wrote:
> You missed the last point - we work on the libraries because we have
> customers that use them.
Yep, now I get it. You must be able to license your code to your
customers using non-GPL license. I hadn't got that point, so I ended
up wasting a lot of our time :-(
>> If we could *selectively* adopt Classpath, it would be great.
> What 'selective' bits of Classpath did you want to use anyway?
Whatever I found to be useful to fix problems, say, copying methods
and classes that we don't implement yet, etc. But now this discussion
is meaningless, because I fully understand (I hope :-) why Kaffe
cannot adopt Classpath.
> I don't think it's minor when a program some of my colleagues
> run on JDK fails to run on Kaffe. :-(
> Then we fix the bugs - what so wrong with that exactly?
Something that's wrong is that I tell them to report the bugs and wait
for a fix or try to debug them themselves. Unfortunatley, most of
them don't have the skills needed to debug Kaffe, and they can't just
wait for a fix from someone else because they've got deadlines, so I
usually end up having to lend them a hand even when I've got tight
deadlines ahead. Although that's good for Kaffe in general, and even
for me in a long term, it's not very good for me in the short term,
when it happens :-(
> In the time I've been doing Kaffe I've had numerous offers of
> libraries, help, ports and so forth - and almost without exception
> absolutely nothing has come of them - possibly this is my fault but
> I'm not sure how to manage this better (while running TVT as well).
I've already talked to Godmar, in private, about your common lack of
responsiveness. Many times, when I started hacking Kaffe, I felt as
if you didn't care about what I was proposing because you just didn't
reply, or didn't sound very glad with the ideas. But now I see you
really don't mean any of that, it's just the way you are, it's not
your fault. It becomes much harder to do something, at least for me,
if I don't feel other people will like it.
I'm not saying you have to spend all day writing e-mail with `Cool',
`I love that idea' or `This improvement would be welcome' (we all can
do that :-), I'm just trying to explain why I, and possibly others,
have non contributed to Kaffe as much as we could. Now there are
several people in the Kaffe Core that can welcome new contributors,
and we can share the load of this kind of personal management with
you. But it would be nice if you could still take some part of that,
because people like to know that the ``king'' :-) is paying
attention.
Just my R$ 0,02 (no dollars, sorry, I'm saving :-)
> Testing is a big issue and something we've addressing here - we recently
> hired a guy to do QA for us so we've hoping to improve this.
Great! BTW, I was wondering whether we should convert all our tests
to the Mauve framework and start testing Kaffe with Mauve regularly,
instead of, again, duplicating the effort.
> Now that you mention serialization, one of their important
> decisions was to implement classes that were
> serialization-compatible with Sun's implementation (but that
> code that may be untested too). We'd have to spend quite a lot
> of time in re-doing what they've already done :-(
> Define a lot of time - I think I've spent more time on this conversation
> than I did rewriting serialiazation :)
As you probably already know, compatibility with JDK in terms of
Serialization is not just a matter of the wire protocol: default
serialization won't do for classes that don't have exactly the same
structure, so, in clean-room implementations, any Serializable code
that wishes to remain compatible with Sun's must explicitly define the
class serial ID and implement read/writeObjectState methods, writing
the fields exactly as specified by Sun. I believe Classpath has
already done that, based on their previous discussions on this
subject, but, again, I may be wrong.
> Argh! GCCs back end as JIT - my god the tradeoffs are so wrong I hate to
> think how bad that could be.
Yep, I know it would be ridiculous to use it as the only JIT, but I
was thinking of using it as a background JIT, with a cache, keeping
current JIT (or even the interpreter) to run a method's code before
the background JIT does its job.
> Actually I should just fold in the GCJ stuff I have and see how well
> it works - I need to update it (perhaps I should do that tomorrow).
Great!
-- Alexandre Oliva http://www.dcc.unicamp.br/~oliva aoliva@{acm.org} oliva@{dcc.unicamp.br,gnu.org,egcs.cygnus.com,samba.org} Universidade Estadual de Campinas, SP, Brasil
This archive was generated by hypermail 2b29 : Sat Sep 23 2000 - 19:58:05 EDT