12 July 2012

Slice and Dice

Seeking Alpha is, mostly, a stock touts paradise. Every so often one comes across an article which is trend seeking, rather than stock pumping (or, as the term is used, bashing). Today is such a day. This article explores the trend in internet connections from the point of view of ads. I don't much care about internet ads; I avoid them with AdBlock Plus. You should, too.

But the article, reading between the lines, makes my case again for high NF datastores. Just as there is less real estate for ads, there is less real estate for user data. In order to maximize the utility of phones, the data has be amenable to fine dicing while remaining coherent and meaningful. Having been embroiled in a few attempts to take VSAM data, conveniently stuffed into DB2 tables, and disassemble it into bite size pieces, I've decided I won't do that any more. Get it right the first time, as Mike Holmes counsels.

Dr. Codd's goal in devising the RM was to define the minimum cover of a datastore (here is presentation which explains the RM that way; see slide 10 and following). In 1969, when he first presented his findings internally at IBM, IMS was the new kid on the block and on its way to being a cash cow. It was not, however, a very coherent datastore. Just as the lawyers at IBM had devised GML (which later morphed into SGML and still later into html and xml) about the same time, IMS was an ad hoc answer to the general problem of BOM processing. Codd saw the flaws and changed the game with the RM. The Suits in Armonk were not pleased, and Larry Ellison grab the ball and ran with it.

In the 1960's DASD was still something of a new idea, and rather expensive compared to 9-track tape. COBOL was coming into wider use, after assembler, and was built on a file/record sequential processing paradigm. Stuffing all the "related" data into as few files as possible simplified, in a perverse way, applications. Applications of that era, to use today's terms, were siloed.

Fast forward with Mr. Peabody to today, and the demand for smaller, coherent data interaction will only increase. If that sounds a bit oxymoronic, well I won't argue. Think of feeding a baby versus Jabba the Hut.

No comments: