Re: Webbed MOO

• Date: Mon, 20 May 1996 20:38:27 PDT
• From:
  Michael Brundage <anomaly@u.washington.edu>
• Cc: moo-cows@parc.xerox.com
• Content-Type: TEXT/PLAIN; charset=US-ASCII
• In-Reply-To: <1.5.4.32.19960521024200.006a7b98@seqaxp.caltech.edu>

On Mon, 20 May 1996, Eric Mercer wrote:
>...webbed MOOs got a reputation as being laggy.  They aren't.
>  Also, the Sprawl used
> to be very populated, and having 100 connected users will lag your MOO more
> than any web system.  Then again, some implementations have more potential
> for lag if your MOO ever gets a few hundred web hits at once.
> Implementations where every web transaction creates a new MOO object are
> probably most vulnerable to this, since you're talking about a lot of server
> overhead.
[...]
>I look forward to hearing a slick rebuttal soon.  :)

No slick rebuttal, just a few comments (which probably go without saying
anyway, but...):

First of all, you can cut down on overhead by caching objects ahead of
time; then connecting via web is just as easy as connecting as an ordinary
player.  If you've got a few thousand users, you can even go snazzy and
put your user data somewhere outside the MOO and swap it in when needed
(though obviously, this comes at a cost of greater overhead).

Second, you always have full control over a MOO -- if you don't want 100
simultaneous Web users, code a cutoff just like some ftp servers (Sorry,
there are already 3 users connected at this time.  Please try again
later.)

Finally, just as a sort of overall comment on MOO speed as a Web server,
you can do it so that MOO is plenty responsive.  When I used MOO as a
gateway to a Mathematica process, I never achieved the speed the
Wolfram folks have on their Web Mma interface but it wasn't much slower
either -- and I was making a separate network connection, plus
waiting for Mma to churn out 3D graphics, plus creating the
HTML documents themselves on the fly (involving a lot of string and
list operations in the MOO). In my opinion, MOO is a fine vehicle for
providing nifty Web services.


michael

• Re: A good set of generic questions...
• From: Eric Mercer <mercer@seqaxp.bio.caltech.edu>

• Prev: Re: verb_info()
• Next: Webbed MOOs
• Index: MOO-cows
• Thread: MOO-cows