on GUIs and such things

[Jonathan S. Shapiro]

> I was curious too... acc. to
> http://www.eurotools.org/Sigs/Metacomputing.html
> it's basically a virtual machine made from a network of other machines,
> but in such a way that each task is distributed to the machine which
> is best equipped to work on it.  Whereas the NOW approach works best
> with machines that are all equal, I guess.

Metacomputing is best thought of as an outgrowth of the "Network of
Workstations" idea. It grew out of the realization that the NOW approach has
some challenges that need to be addressed as you scale up:

    failure of nodes and recovery from this
    consistency across the collaboration
    cost of moving/setting up computations
    differences in communications bandwidth
    specialized resources
    security (for guest computations).

All of these are essentially resource management problems, and if you twist
your head the right way you'll find that most of the papers on metacomputing
are about resource management in one form or another. Security issues are a
late-comer to the problem space (as usual, sigh).


shap