> Niclas wrote:
>> I still thing we need a mini LW of some sort. Maybe if we join
>> forces between DM, VNet and bang we could do an BOMU specification for
>> this now?
Robert Bjarnason wrote:
> Persistent, shared MU is unknown territory, many possible routes
> and I think we should maybe set ourselves small goals with short
> implementation cycles and kind of write this BOMU spec on the fly
> as our experiments become success.
I agree, mainly because I really doubt the ability of the current
crop of VRML clients to handle anything really ambitious. VNet tries
very hard to be friendly to all the VRML viewers, and although that
might limit it in some ways, it means that it's useful right now
to a potentially wide audience.
Also, I read the LW spec once, and it confused me, and that makes
me angry, so now I hate LW. (I don't think I really mean that,
but I'm not entirely sure)
One of my preferences (I don't remember who I stole it from) is to
keep as much of the functionality in the server as possible.
That keeps the clients very, very simple as avoids all sorts of
problems with unreliable browsers and widely varying client
capabilities.
I was thinking about it last night (studiously avoiding the real
work I was supposed to be doing) and realized that you could avoid
the state storage problem I brought up earlier by having the bot
also be the http server that serves up the bot's wrl: that way,
when new people join, the bot/http server gives them the latest
model with the latest state. a cleverly designed bot could give
you shared objects with only very minimal changes to the current
client and no changes at all to the server.
I know I'm harping on this, but you really don't want to hack
the client too much if you can avoid it: The connection to the
EAI is like a house of cards, and even innocuous seeming changes
can mean it suddenly stops running on some obscure browser someplace.
Even if it works it means a round of testing (it even has to
be tested on macs, and who has one of those anymore? :-) )
I guess this means I have to post the client changes I made to
distribute click events. I will pull up the code and diff it tonight
or tomorrow. They seem innocuous enough (see note above)...
Aside: Nicals, Jeff said you weren't on the list? Or were you
just posting from a different address? I have this thing about
keeping the cc: list small, so if you're still on the list I
won't forward any more duplicates to you...
-cks
