Re: Can we solve this -- NFNF and non-1NF at Loggerheads

Dan wrote:

> ...
> The value is a value in the context you discuss it is returned, be it a
> "a character, a set of characters or a set of sets of numbers, or a bag
> of ordered characters". Does the phrasing 'contains a single value'
> still hold true if we want to get a value, open it up, traverse across
> a set of attribute/sub-domain pairs and their associated values,
> continue traversing through nested levels to meet the specified but
> arbitrary criteria, and then retrieve a value (which is actually some
> component of a value), whose domains comprises the definition of a
> higher-order domain or a hierarchicaly ordered set of higher-order
> domains?.
> ...

if i'm reading the several conventional authors (Darwen, Date and so on) accurately, all that 'get'ing, 'open'ing, 'travers'ing, 'continu'ing and 'retriev'ing wouldn't (or shouldn't) happen at the same level of abstraction as the 'schema'. in practice, i suspect all that manipulation is the direct as well as the indirect cause of much, maybe most, of the code in many real applications.

i presume the intent of these authors is that real dbms's would have neat and tidy ways for us to define 'user domains', including the necessary operators (which might be as 'opaque' to a dbms as the structure of a domain's value while staying transparent to the user who invoked them from the dbms).

if i can make an aside, personally, when it comes to RVA's, i think they are an interesting case because one immediately wonders why a dbms couldn't recursively handle them (that is, delve into them at the same level of abstraction as for the containing relation). maybe somebody has come up with an artful way to do that and i just haven't heard of it. if so, could other decomposable 'values' be susceptible to the same technique? for that to be possible, i'd guess that they'd have to be decomposable (or may i say parsable?) using the same structures inherent in the dbms (sets, fd's, headings and so forth). my theoretical CS knowledge isn't strong enough to prove whether, for example, a regular expression could be defined based on a model's inherent structures without having to give up something else. even if it could, what about things that aren't parsable by regular expressions, such as most of the imperative languages? there'd still be classes of things, such as program text that wouldn't be susceptible to such parsing. i'd call them PVA's (program valued attributes)!

i doubt if all this answers the real question, which i haven't been able to think of yet. if i had to write any more on this subject, maybe i'd be well advised to stick to the subjunctive case.

pc Received on Tue Feb 08 2005 - 22:54:03 CET