[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [moca] Locations?

At 9:59 +0100 10/07/03, Martin Berger wrote:
> > The question now is: isn't this just pi? where is the distribution?
>
>it would not be a bad thing if you could show your located calculus to
>be expressible in some pi-calculus. designers of distributed systems
>would love being able to mask distribution completely, alas they can't.
>their problem is essentially that one cannot hide time and certain
>types of failures. since your calculus features neither, it would not
>be too surprising if distribution was not observable. on the other
>hand carbone and maffeis (express'02) do get separation results
>for what may be seen as a distributed pi-calculus, despite
>the absence of time and failures. it also seems unlikely that the
>full ambient calculus is syntactic sugar on top of some pi-calculus.
>this leads me to ask: exactly what feature separates locations (however
>understood) from vanilla name passing interaction?
>
>martin
>


Failures are a crucial point. Modelling fail-stop failures means the
ability to stop an executing process, regardless of its current state. This
is exactly what is done in Amadio's pi1l and in the Distributed Join with
failures.
This can also be done in the Seal calculus with its 'move & replicate'
primitive (you just make no copy of the seal), and in the M-calculus with
its 'passivate' operator. Modelling fail-stop failures with recovery, means
the ability to stop a process execution at any point in time and to restart
it later on. It is possible to simulate such failures exactly with the
M-calculus (passivate as a thunk, then reactivate the thunk by a dummy
application).
Other forms of failures can also be modelled (eg omission failures) but
more complicated forms of failure require the addition of time.

What distinguishes a distributed system from a non-distributed one is that
you can have parts of a computation failing (in various ways). Calculi can
only claim to be 'distributed' in my view if they allow you to model at
least some form of location failures.

Regards,

Jean-Bernard


  
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
The "models for mobility" mailing list     mailto:moca@xxxxxxxxxxxxxxx
 http://www-sop.inria.fr/mimosa/personnel/Davide.Sangiorgi/moca.html