Re: Temporal operations

"Bob Badour" <bbadour_at_golden.net> wrote in message news:<Gc26b.548$rE7.51854452_at_mantis.golden.net>...

 [ snip reference to my quoting the D&D diktat ]
> Brown is taking a comment out of context. The above quotation refers to any
> perceived necessity to extend or to correct or to subsume the relational
> model to provide support for user-defined types and type inheritence. By
> combining orthogonal generic interval operations with generic relational
> operations, the U_ operations arguably extend the relational model, but they
> only do so as shorthands for things that are already valid within the
> pre-existing logical data model.

  Bob - let me quote the entire fucking context.

   "Now, we have said that we acknowledge the desirability of supporting     certain features that have been much discussed in more recent times: to     be specific, features that are commonly regarded as aspects of     "object-orientation. However, we believe that the features in question     are orthogonal to (i.e. independent of) the relational model, and     therefore that the relational model needs no extension, no correction,     no subsumption -- and, above all, no perversion! -- in order for it to     possible to define some language that:

      o Is truly relational (unlike SQL);

      o Does accommodate the desirable orthogonal features;

      o And, finally, can represent the firm foundation we seek."

    Taking the quote out of context? No reference to user-defined types,    or inheritence. Two whole paragraphs later we find the first reference    to them, after the introduction of 'D' and an outline of the methodology    used to frame the manifesto. The diktat refers to 'the relational    model' as the object of the sentence, and the list of points following    explicitly say 'truly relational'.

     Bob: it would appear either that you are a little lacking in    'intellectual honesty', or that you continue to struggle with    basic written comprehension.

    BTW, D&D in the rest of their book promptly *extend* the model with    domain inheritance (not a concept from mathematical logic and, as D&D    themselves say, 'independent of' to the relational model), *correct* the    model by introducing a distinction between relations and and entirely    new thing called a relvar, subsume the core of the relational model    (33 prescriptions) within an edifice that includes 26 other    inessential rules. Does this amount to a perversion? Hard to say.    'Perversion' is in the eye of the beholder: the word lacks a logical    foundation.

     But to broaden the theme: why is it illegitimate to 'extend' or 'pervert'    the relational model with 3VL but OK to add 'domain inheritance'    (OO Prescriptions 2 & 3)? Why does a description of a logical model    include two Prescriptions about transactions? (RP# 17 and OO # 5)?    And if (as they may) D&D remove transactions from their view of the    model does this constitute a 'correction'? Using words like 'pervert'    amps up the rhetoric but adds nothing of substance to the argument.

     Leaving aside for a moment the point that if the U_ operators are    shorthand for simpler operators why go to the trouble, let me simply    point out that the "shorthand" depends for its definition on the    existance of an implicit relation. Every other relational operator can    exist completely independent of there being even a single relation. Not    so the U_ operator (or PACK). In other words, DD&L are introducing into    discussions about the Relational Model, for the very first time, the    concept of a 'compulsary relation'. "except by us!" indeed.

     At what point does it become time to whip out Occam's well worn razor    and accuse DD&L of 'multiplying entities unnecessarily' and invoke the    other D&D diktat to the effect that formal models ought to be as simple    as possible but no simpler?

     There are alternatives. Why not recognize that 'value identity' is not    the same thing as 'value equality' yet 'value identity' fulfills all the    requirements of the other aspects of the model (functional dependencies,    set algebra, basis for proving operator re-ordering rules, etc). This is    how Codd deals with the question in RM/V2[1] (in RT-1). Another question:    why is D&D's approach to view updateability preferred to Codd's (or to    Dayal & Bernstein's, or Keller's)?

    Doing so means that we don't need to suddenly add to the model    propositions like 'The following relations *must* exist", and makes the    'Relational Model' capable of capturing a much broader range of predicates.

     So: I accuse DD&L of violating their own diktat in a situation where    a perfectly reasonable, simpler and equally rigorous alternative    existed. The Relational Model needs no extensions, corrections or    subsumption. I would also do away with domain operators, domain    inheritanc (the whole ghastly edifice of the POSREP) the U_ operators,    and a host of other cruft. All you need are relations, domains, operators    and constraint predicates.

      Nihil curo de ista tua stulta superstitione.

      KR

              Pb

  [1] Codd, E. F. "The Relational Model for Database Management: Version 2"

      Addison-Wesley. 1990.

  (Note that this work pre-dates the publication of the Manifesto by a whole    four years, pre-empts many of TTM's novelties (though not all) *and*    anticipates ideas implemented subsequently by several research &    commercial systems available by the time the TM was composed. Codd's    book also contains its fair share of fluff but it's a work that's been    completely overlooked and should not be IMHO.) Received on Sat Sep 06 2003 - 22:53:39 CEST