Re: Isolates again?

Hi Phillip,
                The delay of this reply is inversely proportionate to 
the probability of my further public discussion of this subject... :-)

You wrote:
> Funny how this came up as a side discussion at a recent Tri-JUG 
> meeting, and now it's a topic of discussion on DZone.
>
> <http://www.dzone.com/links/should_isolates_be_given_a_second_chance.html> 
>
>
> So what do you guys think?  Does Java need Isolates?

As you know, I was the original spec lead for JSR-121 (the "Isolates" 
API).  Some of this stuff you know but I thought the detail might be 
interesting to the list.

Isolates, as the API designers envisioned them with at least three 
implementation "styles", are not a simple abstraction to slap on the 
side of Java like a logging API or the like. Unfortunately, from my 
perspective, we missed the window, which was wide open in 2001 but then 
slammed shut (on our fingers), leaving our collective perseveration to 
run its course until the final, wimpering end when a dip-shit-trivial 
reference implementation of the "essential subset" of the JSR-121 APIs 
was rejected for inclusion in Java SE 6 in December, 2004. Sun Labs went 
on to push 121 to completion with the earlier "first public review" spec 
and a heavily license-restricted "multiple isolates per JVM native 
process" style (i.e. "MVM") reference implementation was made available 
in 2005. I haven't been in touch with Greg since he went to Google and 
don't know anything about the status of this JSR since I resigned from 
the expert group.

But viewed positively, maybe the next evolutionary step should be both 
"beyond Isolates" and at the same time simpler because 1) Java 
implementations have made some very hard problems seem not so hard now, 
and 2) Much of the JSR-121 spec baggage has not likely stood the test of 
time wrt to value (but I can't say for sure).

What I'm trying to say is that adding an abstraction as basic as, say, 
threading, is a large undertaking and the Java community members  (and 
especially engineers working on Java at Sun) seem fully booked with 
other priorities. Viewed from a more "realistic" perspective, Java 
researchers "routed around" JSR-121 a long, long time ago (SAP's 
"Unsinkable Java" research implementation is a good example of how far 
things went concurrent with the JSR activity, as do various "strong 
mobility" projects). But the lack of something available via Java SE or 
ME has been a huge dragging anchor, IMO.

But if you'd like to see unofficial, Sun private API, under the cover 
"Isolates" in action (and moved beyond the original domain limits of the 
JSR expert group to genuine "strong mobility"), notice the paragraph 
starting "Sun SPOTs can be tied together..." in this article:

  http://www.sdtimes.com/article/LatestNews-20080115-11.html

These are Isolates as implemented with the Sun Research "Squawk VM", a 
current research platform (hope I spelled that right).

Another way to put the article paragraph mention of "migrating a running 
application" (as one of the Sun SPOT engineers would put it) is "drag 
and drop a freeze-dried Isolate". 

To quote the VMWare beta test T-shirt from years ago: Powerful Stuff!

Regards,
Pete

PS More details about JSR-121 are available here:

  http://jcp.org/en/jsr/detail?id=121

and here:
  http://bitser.net/isolate-interest

PPS Current Sun management, with about two exceptions, would have no 
clue whatsoever about what I'm talking about here, but I'm obviously 
speaking for myself!
>
>
> TTYL,
>
>
> _______________________________________________
> Juglist mailing list
> [email protected]
> http://trijug.org/mailman/listinfo/juglist_trijug.org
>