Re: Kaffe & Classpath

[Tom Tromey]

>>>>> "Ian" == Ian D Stewart <address@hidden> writes:

Tom> What perception precisely are you talking about?

Ian> The perception that the development of Classpath is driven by
Ian> Cygnus' (or RedHat's) business strategy, as opposed to technical
Ian> merit.

It is true that Red Hat's business has an effect on Classpath.
There's no doubt about that.

I hope that those of us who work on Java at Red Hat (and there are
only four of us) haven't prioritized Red Hat's business interests over
technical merit.  I certainly haven't done it consciously.

My personal goal, and Red Hat's goal (inasmuch as Red Hat has one --
really it is just us engineers), for Classpath is to make it as good
as possible.  There are many aspects of "good" -- complete, robust,
fast, bug-free, compact, readable, etc.  And, largely, I think we've
managed to stay on top of these goals as we've slowly merged code
between libgcj and Classpath.  I don't think I could point to any
piece of code in Classpath and say "this code is good for Red Hat but
bad for everybody else".

In sum I disagree that these two things must necessarily be in
opposition.  Sometimes they might be -- and at those points I think it
is clear that Classpath's goals must hold sway in Classpath's CVS
repository.  (Red Hat can always maintain local changes if they are
required.)

Ian> One of Nic Ferrier's arguments for choosing Classpath over Kaffe
Ian> is that the development of Kaffe is driven by Transvirtual's
Ian> business goals, and that Classpath is not driven by any business'
Ian> business goals.  I don't believe that is true, or at least based
Ian> on past decisions, that doesn't appear to be true.

Do you mean decisions aside from the licensing (I address that below)?
If so, then let's reverse those decisions.  We can back out any change
that goes against Classpath's technical goals.  I'm in favor of that.

So while we might disagree on Red Hat's influence, perhaps we can
agree on goals and consequences.

Ian> Consider, the GPL, more than any sort of methodology or theory of
Ian> computer science, is the hallmark of FSF software.  In the case
Ian> of Classpath, the FSF has chosen to modify what is, at least in
Ian> their mind, a crucial element of the end product for the sole
Ian> purpose of supporting RedHat's business plans.

I honestly don't know what the FSF's purpose was in changing the
Classpath license.  I do know that Red Hat lobbied RMS to change the
license.  Exactly why he made this decision, though, I don't know.
Perhaps he was swayed by Red Hat's arguments about the desirability of
acquiring libgcj.

Red Hat isn't the only party involved in the licensing, though.  The
GCC Steering Committee is also involved.  My understanding is that
they made a decision that runtime (target) libraries had to be
licensed on terms usable by embedded developers.  Now, Red Hat
certainly had an influence on that decision as well.  But I doubt Red
Hat was the only party involved in making it (I'm not on the
committee, so I don't know for certain).

This decision by the Steering Committee affected the licensing of
libgcj.  It also affected Red Hat's decision to assign copyright of
libgcj to the FSF; at the time the SC (or was it RMS?  I forget)
required assignment of runtime libraries.  (Nowadays that isn't
necessarily true.)

If you want Classpath to be pure GPL, or whatever, you're free to
lobby RMS to make that change.  I would disagree with it, but in the
end it isn't my decision, or Red Hat's decision.  It is RMS'.

Tom