Re: WEB : Mapping out communal cyberspace

Mark Waks (justin@dsd.camb.inmet.com)
Mon, 13 Jun 94 13:26:20 EDT


Well, we do seem to have hit on a fundamental issue here (the question
of "fixed space" vs "hyperspace", however you want to call it). I prefer
the unfixed system.

Consider: one of the great strengths of the Web is its enormous
flexibility. The very fact that things are *not* locked into a single
rigid hierarchy makes the Web oodles more useful, because it means
that the odds of your actually finding the data you are looking for
goes way up. A given page isn't found via one single path -- it can
be found via dozens, or hundreds of paths, each for their own good
reason. The result is that wandering around in Mosaic is far, far
more productive that doing so with FTP ever has been.

The key issue is really navigation. A fixed-coordinate system, mapped
by a central authority would be very easy to navigate in -- but
generally much *harder* to find things in. (If you believe that it's
straightforward, or even possible, to organize the whole world of
possible topics into a single sensible hierarchy, you've paid little
attention to the history of Usenet.) The other alternative (which I'll
call "localized-coordinate") is significantly harder to write, but
far more flexible. In the long run, it'll be a win.

So how do we navigate in a localized-coordinate system? We want a
model that allows us the flexibility of hypertext, but doesn't
break the "reality" of our virtual reality too badly. In particular,
it shouldn't be *jarring* as someone wanders in cyberspace. Let's
take a practical example.

Say you've got a site that is providing a technical library, and
you want it to be up in cyberspace. There are at least a couple of
different "streets" that you want this "library" to be on -- for
example, one street that provides a concentration of libraries, and
another that is in a sort of virtual Silicon Valley, a road with a
lot of electronics shops on it.

Now, say that you're a user. You should be able to walk in the front
door, poke around a little, and walk out again -- onto the same street
you came in on. This is the advantage of the fixed-coordinate system:
simply by the nature of the beast, you *will* wind up in the same place
exiting as you were when entering. But we can deal with this problem
in a localized model fairly easily, using a mechanism analogous to
Mosaic's "back" button. Mosaic keeps track not only of where you are,
but where you've been, and allows you to unwind that path easily. Our
cyberspace could do a similar thing implicitly -- when you "exit", you
come out at the same place you went in.

Of course, our library may want to have exits specifically designated
for particular streets -- for example, it's quite likely that we want
to have a way for users to *explicitly* get to the Street of Libraries.
The implication, I believe, is that we want a model with "entrances"
and "exits".

A given "room" defines some number of entrances and exits. Each exit
is a doorway to an entrance in another room. When a user walks through
an exit, he winds up in the other room, and if he walks back through
the entrance he came in through, he winds up back where he started.
However, an entrance simply doesn't go anywhere if you didn't come in
through that entrance -- the door is locked, as it were.

This sort of model should lend enough constancy that it isn't going
to mess with anyone's head -- wandering around will generally produce
intuitive results, while preserving Mosaic's ability to have infinite
links to a single "place". Obviously, the model I describe here is
just a very first step, and needs lots of fleshing out. But I believe
that it's a base model that serves our needs quite well, and we'd do
well to reflect it in the language.

Note, BTW, that there is nothing in this proposal that *prevents* one
from having subspaces that are rigidly organized -- it should be
straightforward to implement a universal-coordinate system under this
local-coordinate system. You'd just need a body that kept all of the
constituent local spaces consistent. (Actually, to make it work
*well*, you'd need to add a mechanism for entrance/exits, so that an
entrance acts as an exit to a particular location if the user didn't
come in through it. That's a small tweak, though.) Given this, I see
no reason to *limit* ourselves to a fixed-coordinate system...

-- Justin

Random Quote du Jour:

Re: February
"Ah, sweeps month. When the network TV execs come out of their holes,
look at their ratings, and look for a way to pander to their audience in
an even more explicit manner than usual."
-- Moriarty