[Date Prev][Date Next][Thread Prev][Thread Next][Author Index][Date Index][Thread Index]
Circular Multi-Orgls & Logic Programming
- To: <mark>
- Subject: Circular Multi-Orgls & Logic Programming
- From: Eric Dean Tribble <tribble>
- Date: Mon, 23 Oct 89 16:31:12 PDT
- Cc: <michael>, <xanatech>
- In-reply-to: <MarkS.Miller'smessageofMon>,38 PDT <8910232311.AA06139@xanadu>
What about loops that are closed only through two backends? They're
guaranteed to cause trouble when you suck the bight into your local
machine. How can you reject them then? How can you prevent them
without making informs wait on a poll of (potentially all) backends
(including archives and down machines)?
I strongly suspect that proper consideration of Bebe, especially in a
mutually mistrustful distributed system, will force us into solution
#3. We should still be able to get away with #1 in first product
unless archiving throws a rationally infinite wrench into the works.
Interesting points. I don't think we can prevent circular structures
in BEBE, so we have to cope with them using #3. Archiving likewise,
except that we will control the archiving implementation for first
product, so we can ensure that it doesn't create infinite structures.
Archiving and BEBE transmission of infinite orgls will require some
semantic support in order to describe the structure finitely. Yech.
Balancing within the ent shouldn't be a problem as long as we know
where the circularities are (which I think is straightforward).