jonathan wrote:

>if i may but in on this thread...
>dictionaries are usually stored as trees, that is:
>a -> advocado -> advocate
>would be stored as:
>a -> dvoca|do -> te
>[excuse the notation]
>that is:
>"dvocado" would be prefixed with "a"
>"te" would be prefixed with "a" + "dvoca".
>
>this probably increases the time to access a word
>[but moore's first law flattens this], but provides
>small and compact files. perhaps not as small and
>compact as some of jay's, but that's another story.
>
>i also understood that ken wanted to add the soundex
>code to each word... as far as i remember [from ken's
>explanation] soundex code [words] are always 4 letters.
>these could also, i imagine, be treed as the root
>would cover a bunch of leaf words.
>
>all this goes to say that using a fixed array of
>even 31 chars is over the top.

Fixed. See previous post in this thread.

>perhaps ken is barking up the wrong tree...

Yes, but it's the only tree available to me.

I wish I understood programmatic trees, but they are more advanced 
than my limited ability.

But I gotta leaf it to you for encouraging this sap to branch out.

Ken