The point of relational as I understand it, was to try to create a database where *calculations* were not required at all. The query language could just start spitting out results immediately without the need for any interim work space. Obviously all you have to do is add a SORT to this and you need interim workspace. Of course if you then add just a simple index to the field by which you are sorting, you again need no interim workspace, although you do introduce a truckload of disk thrashing. The issue, to my mind, isn't that Nested Relational has a string that we interpret as multiple answers, in a single field, but rather that we by nature expect to be able to *break* these out on separate lines, repeating the fields that we see as single valued on each of those multiple lines. That's the reason why first-normal form devotees see MV as different. W In a message dated 12/27/2010 6:18:40 A.M. Pacific Standard Time, antli...@youngman.org.uk writes:
On 25/12/10 04:01, Dawn Wolthuis wrote: > Oh wow, it sure is hard not to jump in on this one, but it's Christmas > Eve so I will render quick opinion and hope that the conversation is > still going on when I really have a chance to respond. > > A couple of opinions -- > > 1. Were it not for the standardization of SQL, a majority of BI would > be more easily accomplished using data modeled for and stored in an MV > database over a SQL DBMS. > > 2. The MV model is more the future than is the legacy SQL DBMS data > model. Thinking that the RDBMS model has trumped the MV model makes > sense on the one hand, but the RDBMS model is now on the wane while > the MV model is part of the group of data models that will wax again, > some under the category "NoSQL." So, were I to write new software > (hey, I am working with a team that is doing just that), I would > choose one of the non-RDBMS models (Ok, I chose Pick/MV) rather than > the old-fashioned, restrictive, and so-last-century SQL DBMS tools. I > did choose a database with the fastest SQL implementation against the > MV model, however. Given that for some applications SaaS software will > replace software deployed by other means, user organizations will care > a lot less about whether the DBMS is Oracle or some other big dog. I've very recently heard of Date's "3rd Manifesto" or "how to force an object database into a relational mould" (my paraphrase :-) When are the relational guys going to learn that the reason object doesn't fit relational is that relational is BROKE! And just to keep you in the loop on something else, look for Wol in the "new contributors" to LibreOffice :-) I haven't actually done an awful lot (it's easy to get a lot of commits while not doing much ...) but they need a new database :-) so guess what - I'm planning to write "nf2engine" for them. The problem is family commitments as my wife now has Parkinsons. And I've learnt some more about "why relational is broken" by discussing things with other people on the sqlite list. We'll see. At some point soon relational will implode, the question is when. Cheers, Wol _______________________________________________ U2-Users mailing list U2-Users@listserver.u2ug.org http://listserver.u2ug.org/mailman/listinfo/u2-users _______________________________________________ U2-Users mailing list U2-Users@listserver.u2ug.org http://listserver.u2ug.org/mailman/listinfo/u2-users
