Re: Swarm futures (LONG)

[Nelson Minar]

Benedikt Stefansson <address@hidden> writes:

>Firstly, to the untrained eye it seems plausible that the Swarm kernel
>(i.e. sans GUI) could  be ported to any Unix architecture which supports
>GNU + ObjC

Yes. Some of the probe code might still depend on libtclobjc, but it
shouldn't be too hard to fix that. I'd be more aggressive and say that
porting Swarm to Windows (WinNT, certainly, maybe Win95) wouldn't be
too tough either, assuming that gcc-objc works there.

>The problem remains the GUI, X-Windows dependency & Tk/Tcl + BLT.
>Again to the untrained eye this seems to be a source of most of our
>ever popular installation blues

I've never really understood why Tcl, Tk, and BLT were such trouble
for people. They certainly are - I don't deny that - but when we
started this version of Swarm it seemed to me like they were pretty
straightforward, major packages used at thousands of sites.

>So why not port the GUI to Java? And I mean _only_ the GUI classes,
>not the trusted old Kernel.

Ah, yes. It's an interesting idea. And an idea that recognizes a basic
fact of Swarm - that the GUI is really just a component, seperable
from the rest of the system. When we started Swarm we even talked
about making the interface a completely seperate program, so you could
run your simulation on your CM5 and display it on your desktop Sun.
We didn't go that way mostly because it's too complicated.

If you look at a Swarm simulation, you'll see just how divorced the
GUI is from the rest of the system. It's a seperate schedule, a
seperate library, etc etc. So there's nothing stopping someone right
now from doing a quick mockup of a Java interface. I suspect you'll
want to run the frontend as a seperate process - linking Java and ObjC
together won't be simple.

*BUT*, and this is a big issue, it takes time to do this sort of
thing. It's been awhile since I've been in Santa Fe, but I don't think
the Swarm team there has any spare time at all for projects like this
right now. It's one more cool thing to do, sure, but it's a separate
project.

It might be worthwhile for someone to do up a quick hack/proof of
concept, like I mentioned above. I'm not convinced Java is appropriate
for this use, btw. I do think it's worth trying.