[I7] Library messages

[David Fisher]

Graham,

> As for what goes in the angle brackets, hmm.
> There will clearly need to be an index page
> listing all of the message names known to NI.
> Still, they ought to be predictable in some way.
> Perhaps named after their standard text? E.g.,
>
> message <Violence isn't the answer to this one.>
>
> Or perhaps named after the rules which normally
> issue them? Or even after actions?
>
> message <attacking blocked>

I'll have a think about it ... using the standard
text is appealing, but wouldn't work for similar-sounding
messages or short phrases without much context.

Might it be worth consulting RAIF about this?

> 6. You might also work further on what we need
> to provide in the way of inflections.
...
> "[The noun] weigh[s] more than [the second noun]."
>
> a more appealing syntax than
>
> "[The noun] [weigh*] more than [the second noun]."

There are already rules for [-s], [-es] and [e-s]
(which handle past tense as well); the only reason
for the "[get*]"-style rules was because of irregular
past tense forms.

See "Handling Verbs" in the extension documentation for more information ...

> ...My current state of progress with this is that I am,
> in fact, still on stage (1).

Quick thought: I haven't always kept up with the changes in
I7, so this might already be in the pipeline ... would this
be a good time to also provide NPCs with the ability to do
whatever the PC can do (even if that ability is blocked by
default)? For example, "Fred, give me the rock".

> This handles everything except the "going" report,
> which as you know is pretty multifarious. (I quote
> the source text below.) It can make dozens of
> variations, I imagine, and took just ages to get
> right. It needs to be split up as
>
> If condition 1 then complete message 1
> else if condition 2 then complete message 2
> else if condition 3 then complete message 3
> ...
> else complete message N
>
> but what should N be, and how would you split
> this so that translators and users could tinker
> with the result?

I can't think of a way that doesn't involve a heap of parameters with 
details about what needs to be described.

It makes it a little more complicated that the ultimate printing code will 
be in I6 (I assume that's still the idea?), so you can't use "[...]" rules 
to print things.

I know I'm dreaming, but an alternative (involving a large amount of work!) 
would be to head in the direction of natural language generation, where you 
provide a syntax tree and rules to assemble it into a printable sentence --  
as described by Daryl McCullough in March 1997:

http://groups.google.com/group/rec.arts.int-fiction/msg/554d51bb54d4fa78

(Just thought I'd throw that in there ... It sure would open up a whole lot 
of possibilities for translation. Yes, I know it's crazy ... but NLG seems 
simpler than NLP in some ways).

David Fisher