Re: Event-Oriented Computing

[glen e. p. ropella]

At 06:59 PM 2/4/99 -0800, you wrote:
>The line would have to be above all those things, of course.  Agents
>can't be allowed to run systems calls or directly or indirectly
>control the program counter.  (Unless you have a workstation that's
>not connected to the internet and you don't care if its files 
>are lost.)

Why not?  We already allow system calls.  Is there no way to 
build a kind of virtual machine such that the agent can have 
access to functions that in a normal process would be dangerous
but in here would only crash the virtual machine?  Does what I 
want to do require the complexity of something like the JVM?

>GR> and c)
>GR> publishing a kind of discovery and lookup API for Swarm so that
>GR> users can begin to use it. 
>
>Besides the existing reflection interfaces: Zones and ProbeMaps, we're
>starting work on a high-level directory assistance interface for
>accessing objects across different address spaces.  This could either
>be used with the existing Archiver (the class that associates names
>with objects, for the purpose of freeze-drying or resuscitation), or
>supplant it.

Cool!  So, is this the kind of thing we could muck with (e.g. permute,
delete items from, switch objects, etc)?  Or is it a "truth" thing where
what you see is what's really there?

>Further, it might be useful to make documentation markup in the Swarm
>protocol files available through the same sort of database interface.
>Unfortunately, the level of granularity in the markup doesn't really
>say much more than would be evident from the ProbeMaps (although it
>includes macros and functions, and some expected-type information),
>and it doesn't try to be any sort of formal semantic data
>representation; it's English.

Well, that could be even better, then.  I suppose the manipulation
path would have to be separate from the sensory path (which is 
good and which is not the case with the probes).

I suppose the next thing for me to worry about is the dynamism
of the protocols.  With defobj, we could generate whole classes
(interface and methods), right?  Can we generate protocols 
dynamically?  I.e. what i'd like to do is, while the simulation is
running an agent would poke around in the directory, find methods
she'd like to ball up into a class, construct that class and then
construct a protocol for objects to adhere to, then instantiate
an object.  Will that be possible?

glen

--
glen e. p. ropella                =><=          Hail Eris!
Home: http://www.trail.com/~gepr/home.html  (505) 424-0448
Work: http://www.swarm.com                  (505) 995-0818  


                  ==================================
   Swarm-Modelling is for discussion of Simulation and Modelling techniques
   esp. using Swarm.  For list administration needs (esp. [un]subscribing),
   please send a message to <address@hidden> with "help" in the
   body of the message.
                  ==================================