Re: WWWlogical idea

Al Globus (globus@nas.nasa.gov)
Wed, 7 Dec 1994 12:01:24 -0800


A possible refinement -- I think you've really got two different kinds
of entities here, which you're using the same node type for. I'm not
sure that that's ideal, either from the viewpoint of the Browser or
for human comprehension. On the one hand, you've got the "container"
objects, the groups that you give hints for like "standard-type wall".
On the other hand, you've got "contained" objects, which can be
essentially any object, and for which you've got hints like "place
upper-right".

At one point I had several different kinds of objects -- including the
ones you mention. Creon insisted that the concepts could be unified
and I think this is good. One can easily distinguish between container
functionality and "contained" objects. Container WWWLogical nodes
have other WWWLogical nodes in them. "Contained" WWWLogical nodes
only contain VRML 1.0 nodes. Although having two separate node types
is a reasonable design choice, I prefer the single node type in part because
with the separate nodes types you need to be careful about what kind of children
the nodes have in the two separate nodes case.

relationship. (And for most contained objects, there's no particular
reason for the hints node to be an SoGroup.)

WWWLogical must be a subclass of SoGroup in order to have child nodes.
Thus, if WWWLogical "contained" mode is to have a WWWAnchor as a child,
WWWLogical must be a subclass of SoGroup.

Actually, if we implemented both this and my scheme simultaneously,
the two might synergize well. One thing that my keyword-based lookup
scheme requires is a way to associate a keyword with any arbitrary
node. We might find it convenient to merge my "keywords" with your
"hints", into a single mechanism; it bears some thought.

Agreed. I'm thinking.

Al Globus