Re: foundations of relational theory? - some references for the truly starving

"cmurthi" <xyzcmurthi_at_quest.with.a.w.net> wrote in message news:3F9493FD.5080305_at_quest.with.a.w.net...
> Firstly, thanks, Paul for non-polemic reasoning, and for explaining one
> difference between the way relational and [Pick] dbs's treat data.

Well I was guessing that Bob wasn't making much headway. ;-)

> It's interesting (and surprisingly fuzzy, using "democracy" and lack of
> parent-child constraints as a paradigm,) but no less an important
> theoretical point. From a *practical* viewpoint, having been in the
> systems and applications trenches long enough, using Pick, not sure it's
> important enough...ie, the need in Pick of having to establish a
> master-slave relationship among the data is both convenient and models
> the real world; it does not *necesarily* constrain you from viewing the
> data on a level, "democratic" field, though I will concede that there
> may be efficiency problems. As always, good design will triumph over
> most odds.
>
> It's obvious that there is not a lot of academic writing about Pick; for
> the most part it has been ignored by theorists.

I'm guessing people bundle it in the same class as IMS DL/1, and when they used to bother, took IMS as the archetype on so forgot about other (possibly similar) systems.
P.S. I'm guessing that IMS is still much bigger (in terms of number of users, licence revenue, amount of data stored) than Pick? Is that correct? (ducks)

> As one steeped in theory
> through my academic years, I've never shed any tears over this; but it's
> equally obvious that trying to explain the model to those who look at
> the world through a relational lens is difficult at best.

Well, if anyone cares, Chris Date attempted to educate himself on 'multi-value' systems (using material from the Web) for the article of his I mentioned. However he admits to having more questions when he'd finished than when he started. He even lists some questions in the article. Maybe some Pick advocate would care to reply in detail to those questions. No guarantee that he would be overly interested in the answers however - he would be looking (rightly) to understand multi-value systems in terms of the relational model (i.e. where they deviate, and the consequences of such).

> Others in this
> thread have suggested downloading a copy of a Picklike and using it; I
> suggest that's difficult; using a db without a set of gui/rad tools
> fresh out of the box is a good road to frustration and failure. This is
> not, of course any fundamental criticism of the Pick model, but of the
> marketing effort it engenders, lukewarm at best.
>
> It's late and I no idea whether these are known to anyone or everyone,
> but I remember this, commissioned by Unidata, now an IBM db: maybe it's
> of use?
>
> http://www-3.ibm.com/software/data/u2/pubs/whitepapers/nested_rdbms.pdf
> Abstract
> This paper discusses technical advances represented by nested relational
> database technology. The focus is the removal of the requirement that
> relational databases conform to the first normal form (atomic
> attributes) and the advantages thereof. Such databases are technically
> called nonfirst-normal-form (NF2) databases, but are commonly referred
> to as nested or extended relational databases.
> We explain the differences between traditional relational databases and
> the IBM nested relational databases (principally the ability to nest
> tables and store complex data structures). We then outline the
> implications of this advanced technology for the user of relational
> databases. We conclude by answering questions that are frequently asked
> about nested relational databases. We provide an extensive bibliography
> of published material concerning nested relational databases.

Again I'd have to suggest Date's normal form article as a response to the above.

P.S. I suspect someone did a find-replace of 'UniVerse' -> 'IBM' in an existing paper after we acquired the technology via Informix. Received on Tue Oct 21 2003 - 12:20:33 CEST