Re: Newbie question on table design.

>> In CDC land we called 'em "Copy Decks" cos they were (originally) 80 column punched card decks. I never thought of them as libraries per se as they were not object code - they were functional though :-) <<

You did COBOL on a CDC? Me, too!~ At Georgia Tech. The thing would convert COBOL data types into floating point, then back again and it still ran faster than the IBM hardware on campus.

And, yes, I remember the card decks, color coded and rubber-banded together.

The three of us are going into "Old Fart Mode" now!

>> is that CDC culture discovered interactive development later than DEC culture did. <<

Yes, but it discovered large data sets from the start since it was a scientific machine. Ever see a data cell machine?

>> I'm just about certain that IBM culture discovered interactive development later than DEC culture did. <<

Sure. Punch cards then tapes. The 3270 terminals batched up a screen full of data, you hit XMIT and it went over as if it were a deck of cards.

>> I've got a lot more respect for Joe Celko than do some of the regulars in this newsgroup. But I think he has a big blind spot when it comes to recognizing parts of the history of computing that were beyond his field of view at the time. In other words, "if I never heard of it back in 1975, then nobody heard of it in 1975". It just ain't so. Not for me, not for you, and not for Joe. <<

I worked with DEC, DG and most of the mini-computers in those days. I remember Dibol and DG-Algol and versions of PL/I on Honeywell.

>> The transition from data definitions being sprinkled throughout the source code (think "FORMAT" statements in early FORTRAN programs) .. <<

My first language, FORTRAN II, then IV and then 77

>> .. to data definitions being neatly laid in the form of metadata stored in system tables, and those modeled according to a uniform industry standard, is one of those evolutions that was marked by a whole lot of little changes, rather than one blinding quantum leap. <<

But what the first tool to use the metadata to check production code against those standards? "this variable is not in the dictionary!" or "part_nbr is defined as CHAR(9) and not INTEGER" or "street_addr must be used city_name, state_code and postal_code as a single data element"

Panvalet and the other source code tools were really glorified copy books with timestamps added. The full data dictionary was a major leap in concept and product, not an evolutionary thing -- like the pre-Cambrian explosion in biology.

>> And if Joe thinks that everyone who uses the term "record" is unfamiliar with active data dictionaries and data definition management, then all I can say is that Joe is profiling from ignorance. <<

I profile from experience, since I teach a lot of SQL. I have found this a very, very strong symptom of "file system mindset"; sometimes people really do mean record and not row when their SQL product uses contiguous storage or data is coming from a file system into the RDBMS and they are concerned with a physical problem they are having.

>> Having said all this, I think Codd's 1970 paper has all the earmarks of a quantum leap, rather than an evolutionary step. I suppose I could have chosen the 1969 paper, but nobody ever heard of the 1969 paper (hee, hee). <<

Has anyone collected all of his RM stuff under one cover? It should be easy to do, but everyone will fight over the commentary. Received on Thu May 03 2007 - 17:40:59 CEST