Re: Notification protocol?

Sankar Virdhagriswaran (sv@hunchuen.crystaliz.com)
Sat, 16 Nov 1996 15:25:48 -0500


>I'd like to see notification addressed in a more general context than just
>indexing, so that it can be used for replication as well -- something that
>might apply to indexing, content replication, and RDBMS replication. I'd be
>interested in others' thoughts about the utility of that.
> - Mike

Mike,

Yes, notification is a more general issue than just for indexing systems.
As you say, it can be used for a number of purposes including maintaining
consistencies between replicated systems.

The challenge however is to come up with a notification system that is
simple enough and general enough that it can cover all potential uses. For
example, a notification system that is based on a topic hierarchy will do
fine for news distribution system. However, if one is interested in
implementing full blown replication of objects, one needs to have an object
system in which publish/subscribe concepts are built in.

We have implemented a notification system based on the latter. However, it
is built with Scheme + our own object system that supports run time class
creation and multiple-inheritance with mixins. We are using this system to
build a configuration management + process management system "over the
web". We wrap relational databases (and other information sources) with our
objects and when these objects change, notifications are generated and sent
to other replicas who might be interested in the change. While the "( )"
turns off a number of people, S-expressions are a great ascii based,
recursive protocol for describing objects that have to go over the wire
(for example to maintain consistency). It does not take much resource to
implement a distributed system of this ilk because of the simplicity.

A workgroup was setup by W3C 2 years ago to look into this issue. It has
been fairly dormant for the last 1.5 years.

Sankar Virdhagriswaran p. no: 617 272 9195
_________________________________________________
This messages was sent by the robots mailing list. To unsubscribe, send mail
to robots-request@webcrawler.com with the word "unsubscribe" in the body.
For more info see http://info.webcrawler.com/mak/projects/robots/robots.html