Re: Reminder, blatant ad

dawn wrote:
...
>
> Is there an example with something primitive like VSAM files (indexed
> sequential files) that would illustrate how an application would have
> to change the logical data model based on some physical design
> decision? I'm clearly missing something (lots). I don't recall ever
> having to make such changes.
> ...

Sure there is, if you're really talking about a true VSAM app (ie., just

   programs that use the access method and produce user output without any intervening 'db' layers). File gets too big for disks, convert fixed-length ksds to variable-length esds. Those repeating groups are sure going to change the logical model when you have to invent new field names for them, and re-write your programs,VSAM not having any support for field names. Plus change all the programs that think a record has only one telephone number to check for two or more, et cetera. (Of course, there really was no logical data model to begin with but let's ignore that overwhelming restriction.) Then when random performance suffers, you add another VSAM ksds or rrds to index the esds and all hell breaks loose with your logical transactions. (I believe DEC had a better 'vsam' with transaction support, forget the acronym for it.) Of course, if you were 'advanced' enough to be using IMS, you had the added professional distinction of changing the IMS definitions as well as much of the above. The small competitors such as Cincom actually made this easier, but not by much, so job security was almost as good as in the IMS shops.

When VSAM came out, I told the boss it would never go anywhere, which probably hurt my pathetic career. My reason was an ignorant one as that was before the western world 'normalized' relations with China, not to mention that Codd was little known then. For many years I was wrong, but everything comes around eventually, even if I was right for the wrong reasons.

If Codd is dead and MV is the coming thing, how come it is being compared to VSAM which is even older than Codd's invention?

If the RM is passe, then surely so is VSAM as a direct alternative, or does it instead qualify because it didn't make the mistake of implementing any of the relational operators?

Isn't there a usenet group for low-level access methods?

Since we're so off-the-wall again here wrt database theory, I guess it doesn't matter if I wonder if anybody can recall reading an interview of Codd, perhaps in DBMS magazine fifteen or so years ago, where he said that the term 'normalization' was partly suggested to him by hearing of Nixon's 'normalizing relations' with China? No good reason for asking, I'm just curious (I read CJ Date somewhere saying this wasn't so, but I remember reading that interview.)

pc Received on Wed Feb 15 2006 - 22:39:46 CET