Persistence: a proposal

Kelly Murray kem@Franz.COM
Thu, 22 May 1997 15:58:35 -0700

Scott> I have some AI research I want to do which requires a
Scott> high-performance persistent Lisp with firstclass
Scott> continuations.

Can anyone please explain to me how continuations make things simpler
than traditional threads?  
Or how they can work in a shared persistent OODB?

Patrick> Have you considered RScheme from Paul Wilson's OOPS group at Texas?
Patrick> They have reasonable performance by generating and dynamically loading
Patrick> C. I don't doubt they know how to improve their performance.

C code has been forever banished from my SilkOS..

Scott> I spent a good half of 1994 writing a persistent memory manager/OODB.
Scott> I got maybe 1/3 of the way through before I ran out of money.  

Sounds similiar in progress to what I've got implemented (98% in Lisp btw)
The Brooks real-time GC indirection strategy is interesting,
and similiar to how Smalltalk works and manages to do interesting
things.  But it will slow down the code a lot, increasing paging by
virtue of bigger code size and bigger objects and less locality.
It doesn't seem to be needed either, using VM paging hardware is

-Kelly Edward Murray
[ going back to actually working now .. ]