Re: Let's start afresh

Christopher Browne <cbbrowne_at_acm.org> wrote in message news:<bn2714$s2g8c$1_at_ID-125932.news.uni-berlin.de>...
> The world rejoiced as michael_at_preece.net (Mike Preece) wrote:
> > In providing the list of links for multivalue database downloads I
> > forgot to mention Cache. It's a little different from the rest, and
> > I've never worked with it, but it claims to be
> > multivalued/post-relational ("Pick"-like) too. Here's a page that
> > might be worth looking at in the light of the current discussion:
> >
> > http://www.intersystems.com/cache/whitepapers/performance.html
>
> My understanding is that a BIG part of what Cache does is to require
> that all data be stored in memory, and use algorithms (T-trees,
> notably) that are specifically attuned to in-memory searching.
>
> That (use of specialized algorithms) would be an actual _reason_ for
> differences in performance, as opposed to the vague notions of
> "because of using technology that we are incapable of coherently
> describing."

I was recently pointed to the origins of Cache. It is a descendent of MUMPS and not of PICK. However, jBASE at www.jbase.com, a solid implementation of the PICK model, does have the ability to persist data to an RDBMS as well as Cache (as well as MV). MUMPS was developed about the same time as PICK but is decidedly different, although also permitting nested relations and claiming to be post-relational when it appears to pre-date Codd's 1970 ACM paper.

There are several books on PICK and web sites to learn about PICK if a description is what you would like. I also have what I believe to be a brief and coherent description of PICK in a set of flashcards. It is a "trilogy" with the first set being A MultiValue Overview, the second being MultiValue Data, and the third MultiValue Commands.

In looking over this list before posting, it appeared that this forum was about discussing database theories and that it was typical that the folks in this forum would know at least the models that are used in production in over a million shops, so I haven't thought about writing up a description. But given that college courses in databases these days often don't mention PICK, MUMPS, and other commonly-deployed database approaches, perhaps we should drum up a list of URLs, such as www.multivaluedatabases.com to help out.

Also, feel free to contact me at dwolt_at_tincat-group.com and I can perhaps help with a coherent description (but what are the odds, given my posts to date, eh?), at least at a conceptual level. Others who have authored code making up one of the PICK or PICK-like databases would be better at giving a description of the innards.

Thanks. --dawn Received on Tue Oct 21 2003 - 14:46:17 CEST