[MUD-Dev] Clients

Jon Leonard jleonard at divcom.umop-ap.com
Fri Feb 20 11:55:30 CET 1998


On Fri, Feb 20, 1998 at 10:22:19AM +0000, Vadim Tkachenko wrote:
> Jon Leonard wrote:
[snip]
> > There are quite a few equivalent "machines" characterized by being able
> > to calculate anything that is in theory calculable, while being amazingly
> > simple to describe and useless to program in.
> > 
> > Others include:
> > State machine with two counters
> > State machine with one queue
> > Lambda calculus (Functions of one variable, no explicit recursion...)
> > Transformational grammars (when you see "abc" replace with "def")
> 
> Not quite so for this case, the most noticeable examples are UNIX
> sendmail (which is almost all rewriting grammars) and Eliza robot (btw,
> the funny consequence is that it is possible to implement Eliza robot
> using sendmail - it requires a pervert sense of humor, but greatly
> improves the understanding of this technique ;-)

Assuming you mean "not useless" rather than "not turing complete"...
I'll still claim that without variables or regular expressions,
transformational grammars are amazingly useless to program in.
The version I was referring to above only allows rules with literal
strings being replaced by literal strings.

Expanded versions of some of these languages can be quite usable.  Lisp
and friends are extensions of the Lambda calculus, for example.

> > All of these are far more useful for doing proofs than they are for
> > MUD design, and rather represent things to avoid when doing internal
> > languages.
> 
> Also, what I'm thinking about, is to make NPC engine[s] somewhat based
> on Eliza behavior rules, slightly modified for each given NPC...

Applying it to situations in addition to verbal interactions could be
useful.  "Whenever hit and unhealthy, flee" and the like.

Jon Leonard



More information about the mud-dev-archive mailing list