RE: [cp-patches] Implementing Thread.sleep() via Thread.wait()

[Mark Wielaard]

Hi,

On Fri, 2004-12-31 at 12:10 +0100, Jeroen Frijters wrote:
> Mark Wielaard wrote:
> > Sure if you think that is a good idea. But are you sure we want to
> > encourage arbitrary behavior? I believe we should try to
> > encourage/document consistent and nonsurprizing behavior.
> 
> It's funny, but I'm using the exact same arguments (i.e. consistent and
> nonsurprizing behavior) to defend my position ;-) It's just that I value
> compatibility with Sun higher than adhering to the spec.

And of course there is no real spec. We are just following the different
books published around the standard core classes and try to make them as
compatible were possible where that is in the benefit of us free
software developers. Which is almost everywhere. But in this case it
looks like it is just a stupid bug in at least Sun's implementation. Bug
report filed, expected behavour documented and tests written. Unless
there is a clear advantage to replicating the buggy behavior (some
important free software program depends on this) I suggest we just move
on and hope they fix their implementation.

Last month a couple of us spoke to Onno Kluyt from the JCP (java
community process) and discussed the compatibility question a bit. The
main problem is the NDAs that are some parts of the JCP process specs so
we cannot use them and the fact that the TCK (test compatibility kit)
isn't free software and cannot be used to develop a GPL compatible
implementation. One proposal was to free the TCK and merge it with
mauve. Since with mauve we already have what we want from such a test
compatibility kit and it is already out in the open. Dalibor is working
with Onno on some way to get kaffe certified as "java" since Sun seems
to like that happen to prevent "fragmenting" the developer base. I hope
this will happen since kaffe uses GNU Classpath and their feedback would
help settle a lot of this little issues. But till then (and it seems
that there has been very little progress over the last month) we need to
make the best core class library that we can. And that does not always
mean just replicating some bug in some proprietary implementation.

Cheers,

Mark

signature.asc
Description: This is a digitally signed message part