[alicebot-archcomm] <that> when

[Noel Bush]

> I need more time to look into this problem.  I believe that
> some AIML in the A.L.I.C.E. brain would break if we changed
> the behavior of <that/>.
> 
> Consider a simple YES pattern with <that>
> 
> <pattern>YES</pattern><that>CAN PIGS FLY</that>
> 
> If the client utters any synonym of YES, like
> YUP, UH HUH, OF COURSE, AFFIRMATIVE, and so on,
> then each of those categories (as written in A.L.I.C.E.)
> will contain <srai>YES</srai>.  We would want to
> preserve the value of <that/> when we match
> the category with <that>CAN PIGS FLY</that> 
> 
> Conceivably, there could be a cascade of <srai>'s linking
> a collection of synonyms to the same template, e.g.
> YAH-->YEAH-->AFFIRMATIVE-->YES.

Yes, I tend to think that this is the expected combined behavior of
<that> and <srai>, and that changing this behavior will throw a lot of
existing AIML (not only ALICE) into disarray.  In other words, the
current capability in Program D is just an accident.