Java instead of freedom - the best option?

Tom Thornhill
Tue, 6 May 1997 14:05:05 +0200

Bill House <> wrote in article
> Tom Thornhill <> wrote in article
> <01bc5949$fabf0a80$725e9082@mniw1001>...
> > 
> Well, you're obviously pretty well-versed in VM technology, 'cuz you've
just given a
> pretty good description of one of the VM architctures we're considering
for the UVM.<g>

I'm told TAOS did something very similar ( but I never managed to track
down the 
specs ). So the proposed VM wil have a Mips like instruction set ( though
with stuff like
delayed branches removed )  that will get compiled on install / run. This
has the advantage 
of being a very easy operation ( I reckon you could write a decent native
code generator for such an 
architecture in a month or so - you could make all the branches in the VM
code through 
arrays of function pointers which would be optimised out of the native
code, or something like that ) 

Actually there's probably a VM type that is better than imaginary Risc or
stack based
if you don't ever intend to interpret ( IMHO a good starting point ). Maybe
we should ask
Hennessy or Paterson what they think ;-). The real reason that I like VM's
is that they let people 
experiment with processor design the way the currently experiment with
expansion cards. A truly 
universal VM would let you run an application designed for an API ( whether
Java / Win32 / MFC / Unix ) 
written in any language run on any processor that the API has been ported
to. I think the problem with 
Java is that the VM concept(good) gets bound up with a new and language and
class library(debatable).

> > 
> > But a Microsoft VM would only support the Win32 API, Microsoft have no
> > interest in abstracting the user interface API.
> > 
> Here is where the Java team has done a good job, even though it's a very
hard thing to
> compelling reason to do otherwise.  BTW, part of the UVM architecture is
> multi-lingual, so RTLs that target specific environments, like Win32, can
be easily
> added.

I'm not denying that Java is does a good job abstracting the user interface
I'm saying that it does too good a job from Microsoft's point of view. The
UVM should
let them continue to promote Win32 / MFC if they want to.

> Bill House
> -- 
> Note: my e-mail address has been altered to
> confuse the enemy. The views I express are
> mine alone (unless you agree with me).

PS - I've seen some references on the mailing list to "optimisation tables"
- any one care to
clarify post a FAQ / Spec to the list ( ditto for any Taos info ).

PPS - For god's sake no one tell the guys in comp.lang.asm.x86 about this!
Talkto:Tom Thornhill, Philips Medical Systems Nederland BV
Tel +31 4027 64080 ( Work ), +31 499 371230 ( Hotel - ask for room 19 )
************************ +44 3709 46045 GSM ****************************
**** IMPORTANT: please use the above numbers rather than the GSM *******
****  one if possible as I pay for incoming  calls from outside the **** 
****  Netherlands. *****************************************************