The Mosner bit

Mats Nygren nygren at sics.se
Sat Sep 2 10:09:58 UTC 2000


spair at advantive.com (Stephen Pair) wrote:
> This is very interesting!
> 
> Has this been tried in other implementations before?

I have experimented with it, but it wasn't a Smalltalk system. And it
never reach production, only my own experimentation.

One of the things I did was a lexicon for latin. There are many word
forms and by coding it I had immediate access to millions of word-forms.
Without ever using a hash table or allocate something on a heap. It was
possible to associate meanings with the word forms. Assigning meaning
to the word stems (< 1000 in the experiments) meaning arose for all millions
of words. I used that for quite different programming language with millions
of meanings reachable from a single "identifier". After that most programming
languages seems to lack something. IM(Humble of course ;-)O.

I also had a comprehensive set of primitive functions and that worked
well to.

And other stuff .. as you say below encoding schemes can be useful.

> Does this essentially mean that up to 128 immediate classes can be added,
> each with up to 16M unique instances?

Exactly. 1G + 1G (new tags + smallints) values outside of object memory.

> [snip]
> The ability to efficiently represent just about any encoding scheme is
> reason enough in my opinion.

Good point.

> I think it's definitely worth exploring!  You have my vote.

Thanks!

/Mats





More information about the Squeak-dev mailing list