VRMLWorld.net

[Rick]

Recently, on the www-vrml mailing list, there has been an
on going discussion about features lacking in VRML/X3D.
Many have commented that all they need is a NetworkSensor
node and their multi-user problems would be solved.

While I agree a NetworkSensor node would be a nice feature,
I don't see how that would make 3d chat any better. Just look
at some of the recent efforts to use a 2d text interface in a 3d
world. Typing in 3d just isn't any fun. Even typing in the
3d Second Life interface makes me cringe.

I think, one of the nice things about the way ABNet is designed
is that you type in the Internet Explorer window and not the 3d
scene. By typing in the Internet Explorer window, you don't end
up with lag, even in an animation heavy 3d world. When your
keyboard focus is in the text input field on MSIE, it is the active
thread, not the VRML browser. That means it gets a higher
thread priority than the 3d scene. This eliminates lag. Also, you
can cut and paste from any other application without problems.

If you try to implement the chat interface in the 3d world,
you end up with problems. Say you have a world that is
eating up all your CPU cycles doing animations, it can kill
your ability to type. Your keyboard events are just
another event the VRML browser has to deal with. The
keyboard doesn't get any more priority than any other
event going on in the 3d scene.

If the typing interface is in 3d, how do you cut and paste?
What if I type in a URL? Can I click on it? Will it load in the
current world or launch a new window. What about formatting?
Can I change the color of the text for some of the characters
and not others? Can I bold or italicize it? Maybe, but not
as easily as wrapping some html tags around it.

I can't say I see any improvement to 3d chat by adding a
NetworkSensor node. Why don't they focus on some dynamic
lighting affects, physics or some other things that will actually
make VRML / X3D more appealing.

-rick

[fabricator]

A NetworkSensor node wouldn't just be for Multiuser chat. Imagine being able to connect directly to IRC chat or and industrial process. Create your own custom Shared Events such as whiteboards with shared PixelTexture nodes.

Vrml has GET but not POST, so its hard to send large amounts of data at once, this is one reason there are no online editors for vrml.

Really there are two types of chat.
1) Immersive full screen 3d (secondlife, BattleField 1942, MMPOGs etc) which is where the chat and 3d is all one intergrated enviroment.
2) Multi tasking chat (IRC, blaxxun, vr4all, abnet2), this is where the chat is a seperate system, you can move in 3d, chat, read your email or whatever.

We do need both to be honest, people like reading their email and chatting at the same time.

Most games have the chat as a part of the 3d, but vrml developers haven't tackled that area yet. We do however have a increasing number of chat joined to 3d via html vrml chat systems.

Heck even Islands (Parallel Graphics) used frames and html even though it was a stand alone MU vrml application.

ps: physics in being worked on by the web3d working group at the moment. So far they have a first draft which sounds promising.

[Test Dummy]

From the NetworkSensor draft proposals I have read,
it didn't seem like you would be able to implement
arbitrary protocols. It seemed like the data being
sent was all geared to X3D node types.

http://planet-earth.org/x3d/networkSensor.html

How would you go about implementing a complicated
protocol like the one used by MSN? With vrmlscript
and routes. I'm thinking that would be painful to
debug and maintain. MSN uses a combination of
HTTP, HTTPS and TCP.

http://www.hypothetic.org/docs/msn/notification/authentication.php

Have you looked at the IRC protocol? I have, it really
isn't geared for multiuser realtime. IRC will start rejecting
your posts if you send them faster than 1 per second.
Server operators can make the time between posts
even higher than that.

The suggested implementation of the HTTP protocol in
the draft cripples it. It proposes to use just the HTTP
GET method. I didn't see anything in the draft that lets
you set the HTTP protocol version, provide headers or
even create specific http requests type such as POST,
HEAD, DELETE, LINK, UNKINK, LOCK, UNLOCK.
Will it support persistent connections? How about HTTPS?

Even though it suggests support for UDP, that will be,
useless in a true Internet environment. It would only
be useful on a lan or to communicate with a process
running on your desktop.

Yes all these things are useful but you could implement
all of this today. I think both Cortona and Contact let you
write C++ custom nodes. Why don't the people
proposing these things go off and implement such
a node and then produce a working implementation
for at least those two browser. A working server that
goes with it is also needed. Then a sample world showing how
it all works. Then if others decide it is a good thing at least
people will start from a true understanding of what the
NetworkSensor is and not some arbitrary spec that ends
up getting implemented 10 different incompatible ways.

My real point of my post, was to say that I don't like typing
in 3d. Clicking, moving, navigation that would be yes, typing
that would be no.

-rick