Integrating Kaffe and Mozilla ?
Robert S. Thau
rst at ai.mit.edu
Thu Apr 16 08:23:35 PDT 1998
A few notes.
Tim Wilkinson writes:
> As far as I'm aware there are a number of problems in doing this right now,
> the major one being that the current release of Mozilla doesn't have the
> much talked about OpenVM interface (or did they change that already?).
> Anyway, from what little I know there are two issues in getting Kaffe into
> 1. Interfacing the thread and memory management models. Mozilla has (or
> will have?) a well defined interface to a non-portable OS abstraction layer
> and Kaffe needs to be targeted at this too.
It's called the NSPR layer, and it's part of the distributed source.
It was originally written as the platform-specific support for their
own ports of Sun's JVM, and was later extended to cover other needs of
the software. So, it includes threads, gc (including hooks for
finalization and weak pointers), etc. --- but I don't think it
includes graphics interfacing (see below).
> 2. Pluging the AWT into Mozilla. No my area I'm afraid (Peter Mehlitz
> could probably talk about this) but I suspect it's not trivial.
One complication here is that there are already Mozilla builds based
on two different X toolkits, with a third on the way --- the original
Motif-based Mozilla front-end (FE) code, a port done by Troll Tech to
their non-DFSG-free "Qt free edition" windowing toolkit, and work in
progress to get it running on top of the GIMP collective's Gtk. (I'm
not counting the free Motif clone, lesstif, as a separate port here
since current development snapshots are up and limping with the
standard Motif FE). To complicate things further, Gtk is not based on
the "standard" Xt framework, and I don't think that Qt is either.
Which all can be taken as good reason to base the interface between
the Mozilla FE and the JVM on lower-level stuff, like underlying X
window ID's --- but I'm not sure that Netscape has gone into that much
technical detail on what their "open interface" will look like, and I
*think* I recall Arthur van Hoff saying that the original AWT was
Motif-based at Netscape's insistence.
At any rate, some early coordination with the people who are
presumably working right now on trying to define the "open JVM"
interface might be in the best interest of all concerned. The people
listed as contacts for the Java stubs (OJI) on the Mozilla "owners"
page (http://www.mozilla.org/owners.html) may perhaps be a good
starting point, since I can't find any specific contacts listed on the
"Open Java FAQ" or nearby pages...
(A final note: If you look at the Mozilla community pages, you'll
notice that they've set up a forum for Java issues. However, it's
dominated by people discussing a project to rewrite the whole thing in
Java, and there is very little on the --- to me, more interesting ---
project of getting a usable JVM in status-quo mozilla).
More information about the kaffe