We have just began an activity to investigate how we could interface
our networked (non-DIS) VR tool, DIVE, to WWW. It therefore seems
worthwhile to look at VRML. It's unclear yet whether this would mean
that our tool would qualify as a "browser" or just as being able to
"hook in" VRML specifications just as any other vr format.
Currently, we use IP multicast groups to connect multiple participants
within one virtual world. Each participant may change worlds (and
thereby multicast groups) by entering portals. Worlds are initially
described by a world description format (.vr), but once read, the
state of the world is replicated among the participarting peers, a new
participant gets a complete description from an old peer (not from the
world file). Worlds and their objects are completely dynamic, so they
change constantly over time.
It would be very nice if we could find a correspondence between DIVE
worlds and VRML descriptions. Today, it seems like VRML specs could
be used to some extent, as a template, as an initial description of a
world, although our world descriptions need much more information.
My main current concern is that the DIVE tool is highly interactive,
multiple participants may manipulate any object, define new ones,
communicate through text, sound and images, etc. The peers therefore
need to communicate actively over a multicast address. This seems
unclear to me how it could be solved given current www/http
philosophy. I understand that VRML is evolving and will eventually
incorporate interactivity, and I hope this will be an important future
issue.
(DIVE is a distributed VR tool freely available for academic use by
sending an email to dive-request@sics.se. Current platforms are SGIs,
SUNs and IBM AIX-RS6000s.)
Regards,
Olof Hagsand
SICS, (Swedish Institute of Computer Science),
Stockholm, Sweden