Re: I was thinking about our pepper shaker discussion

Rohit Khare (khare@w3.org)
Mon Aug 26 17:27:59 1996 EDT


1) this is by way of abrupt welcome to two new FoRKees:
- Jamie O'Keefe, BCS/NeXTite, now aiming for an MS in environmental
economics at BU, formerly of Athena Design and First Call
- Louise Woodroofe, only for the next four days, though -- off to Japan
see: http://xent.w3.org/FoRK-archive/summer96/0455.html

2) this is to forward Jamie's thoughts on the "golf-ball network" to FoRK
- I only had pepper shakers handy, so to him, it's the PepperNet...

3) this is to hint at a vision memo I'm working on to be FoRKed soon...

4) this is to note that the summer'96 archives have hit the 600 message
limit, and I've started a new Fall hypermail archive.

okeefe@messaging.tfn.com (James O'Keefe) wrote on 11:32:36 Thu, 22 Aug 1996:

> and thought of a few more quandaries.
>
> The nanobots make the pepper shaker and place it on a table. It randomly
> creates its id and announces its id/presence to all of the local items
> (utensils, plates, room, pepper shaker filler, etc.) Provided it no one else
> has that id it goes on its merry way until it gets low on pepper. It contacts
> the pepper shaker filler and gets filled right? But how does it know it is a
> pepper shaker? Or more probably, who assigns it the type=pepper shaker? This
> seems to be another global assignment problem. You can't leave it to the
> manufacturer since they have an incentive to create their own type system so
> that their pepper shaker understands how to call its own pepper shaker
> filler, but not that of another manufacturers and vice versa. Lock in. what
> we need then is a commonly agreed upon language and a dictionary for that
> language. It would need to be one that evolves, but I am not how it would do
> that.

Ah... but that's what PEP solves: negotiable, extensible, aggregatable
interfaces. There's profitable room for brokers and managers to midwife
conntections between disparate ontologies.

> In regards to the local vs global naming issue, perhaps we should think of a
> group of objects within a certain area or assigned to a certain person as a
> collective. Likewise collectives could contain collectives (i.e. buildings in
> a block, blocks with in a city, cities in... you get the idea.) Then if I
> say, purely hypothetically, display the status of USA:MA:Cambridge:Rohit
> Khare:Kitchen:lights, it would do so. To simplify matters, perhaps Rohit
> Khare is the only one in MA, so we could refer to USA:MA:Rohit
> Khare:Kitchen:lights instead. Then if another Rohit Khare moved into MA, he
> would have to choose a new id. All done locally. Now what happens if your
> Utensils decide to succeed since you aren't treating them well? We will have
> to see. The naming system is still hierarchical, but it is established from
> below, not from above. I can hear those Libertarian pepper shakers now. :-)

Dynamic hierarchy is the key. From below is also exactly right -- that's
where emergent behaviors comes from...

> I don't know if that brings up any useful suggestions, but I'll say them
> anyway.
>
> Jamie

This was a great recap -- I'm really glad that I made even this much sense
after all that Scotch... :-)

Rohit