Let me explain my take on this problem:
One of the parts of my system is an intelligent web crawler. I am
encoding the URLs that the system encounters as facts and reasoning
about whether to persue them or not. The problem is that the queue of
unexplored nodes in a web search grows in size very quickly.
Typically, I'll want to reason about just those URLs that are
associated with a particular page, or just those that point to a
particular host. And, I want the assertions about unexplored URLs to
persist between JESS incarnations, potentially sharing them between
multiple crawler processes on different machines. So a database seems
like a good solution -- that's how real web crawlers work anyway --
rather than having strictly JESS-based storage of URL facts.
I think the backward chaining solution will solve my particular
problem. I haven't had time to upgrade to 5.0 yet, so I haven't looked
at the backward chaining mechanism in detail. However, this solution
does not address the caching problem. You don't want every database
row to end up as an assertion through gradual accretion, any more than
you want to do it by bulk loading.
Perhaps this could be part of what "issue-query-and-assert-facts"
does, namely retract old facts, but you'd only want to retract facts
that you didn't need at the moment. Determining this might be tricky
in general, but in my case, I think it will work. The crawler will
know when it is switching from one context to another and what URL
facts will be relevant in a new context.
--
robin
=:-{)
University of California, Irvine
Information and Computer Science Dept.
http://www.ics.uci.edu/~burke/
"The universe is made of stories, not of atoms" - Muriel Rukeyser
---------------------------------------------------------------------
To unsubscribe, send the words 'unsubscribe jess-users [EMAIL PROTECTED]'
in the BODY of a message to [EMAIL PROTECTED], NOT to the
list (use your own address!) List problems? Notify [EMAIL PROTECTED]
---------------------------------------------------------------------
