Re: ID field as logical address
Date: Sun, 07 Jun 2009 01:08:30 GMT
> IDs. If table header can be viewed as a hidden row, then why can't one
> introduce a hidden ID column? ...
Vadim, I would object to that argument on the grounds that the header is not hidden, otherwise projection would violate the Information Principle. Another reason it isn't hidden is that it is in fact derived from tuples that have triples that have all the attribute names and types that a so-called header has.
Personally I don't object to hidden ID 'columns' as long as the algebra can't refer to them, if it did then they wouldn't really be hidden. Nor do I object to generated ID 'columns' as long as views can reference them and as long as views can have a generated ID even when the referenced base 'tables' don't. I wouldn't prescribe any particular content for generated ID's either, as far as an algebra is concerned, it should make no difference whether they are physical addresses or sequence numbers of some sort. If users choose to interpret the ID content in some way that the algebra doesn't depend on, that's okay by me. Received on Sun Jun 07 2009 - 03:08:30 CEST