Re: Caching and Common Objects

Peter J. McCann ([email protected])
Tue, 30 May 1995 13:42:35 -0500 (BLT)


Justin writes:
> One of the key elements of the keyword proposal is that it
> *specifically* does not depend on any sort of centralized entities;
> it's designed to work within the functional anarchy of the Net. I'd
> like the above proposal a lot more if we could come up with a variant
> that isn't dependent on some sort of central authority. (And I suspect
> we could do so if we tried...)

An inheritance-based specification of objects does not necessarily imply
any central authority or registry; I was just trying to satisfy requirement
#3 in the framework of inheritance. You could just as easily do a local
lookup of the type name or ask the original server for an expansion of the
type (as a scene graph with lots more nodes, perhaps).

Perhaps a keyword based scheme provides a bit more flexibility, but it
does not give you a logical relationship between types (inheritance
tree) that could clue you in to which substitutes are acceptable in what
situations.

I just took a look at the final draft of the spec and found the "isA"
field name. This is probably equivalent to what I have proposed, and
seems already established. Maybe it should be expanded with a list
of keywords describing the added functionality of the new node, which
would enable keyword-based searches throught the class library, wherever
that repository happens to be.

-pete

-- 
Pete McCann                                          [email protected]
Department of Computer Science           http://swarm.wustl.edu/~mccap/
Washington University in St. Louis