12 August 2013

VT Phone Home

Well, The Yellow Brick Road has been assaulted by sinkholes and mudslides of late. Should we fight on? Or capitulate to VSAM? I say... fight on.

Here's a (partial; I'm not omniscient, though I try) list of calamities.

SSD makers.
Fusion-io, Stec, and OCZ are all in deep doodoo. Fusion-io can't make its Wall Street numbers, and appears too dependent on Apple and Facebook and perhaps a few others. Stec is now part of Western Digital, so who knows. OCZ still hasn't figured out its numbers. On the other hand, there remains a slew of niche, mostly private, pure-play SSD companies. I got on the mailing list for one of them, SolidFire, and their latest mailing is for something called the SF9010, claiming to "At full scale of 100 nodes, the SF9010 is the world's largest, fastest all-SSD storage platform on the market - available for less than the cost of traditional performance disk solutions."

Samsung has announced a vertical, higher density, NAND. Samsung and Micron, possibly with Seagate and Western Digital, are likely going to out-vertically integrate production of NAND, and dependent devices. Just as Content, by some lights, controls the consumer device market (say goodbye, Apple), so too will the physical investment in hardware control the device market. Apple was happy with the rich 20%, and the strategy worked. Until it doesn't.

The knuckleheads have a head of steam, no doubt about it. But thinking people understand that an application API world is just fancy talk for siloed data. The RM sought to change that, among other things. And data without ACID is just a lump (I typo-ed that as 'limp'; where's Freud when you need him?) of bytes. Those that ignore history are doomed to repeat it. Now, the basis of the NoSql argument is that a database engine can't keep track of all that data to be searched (which is where most of this comes from). So, it's back to dumb files and smart code. So, such coders are forced to write a simple minded TPM, or ignore the whole problem.

Oddly, Postgres folks are well into Foreign Data Wrapper support. This is called Management of External Data in the 2003 SQL, but has been around as Federation since at least 1990. Progress, not exactly a pure RM/SQL engine of course, could do it then. So, there is hope.

The PC is market is dying!! The PC market is dying!! Remain calm! All will be well?

What Apple started, or more accurately continued, was the revelation that non-business computer users didn't really need a 'computer' and only had one because the applications (not the programming infrastructure) of interest ran on one. Apple transformed itself from a computer company, which built its machines (almost) wholly from bought-in parts, to a toy company with ever more powerful embedded controllers. And consumers realized that they didn't need computers because they didn't create computer programs. D'oh!! Lotus 1-2-3 was the first step down the slippery slope; if one ignores VisiCalc. Either way, IBM's notion that the Personal Computer would be used by engineering types to write programs quickly, for themselves rather than wait for the crew running the mainframe to churn out COBOL or FORTRAN, turned out to be terribly wrong. Well, and letting the BIOS loose.

The spreadsheet meme, thanks to 1-2-3, has given us naive' views of data, and the London Whale. Not to mention that much of financial services, across the board, runs on spreadsheets. More than any other factor, the spreadsheet has been the leader of the assault on the RM and SQL.

I see glimmers, ever so faint, that the realization that large format keyboard applications don't work so well on small-ish devices sans 'real' keyboards. Touch means picking, and picking means listable data. Listable data means narrow definitions (tables). And, finally, narrow tables means Organic Normal Form™. Back about 1990, Phil Johnson mentioned in passing that he'd never trust a schema that was shallow and wide; a few tables with lots of columns in each. He wanted to see lots of tables with a few columns in each. He never mentioned the RM or Codd or the like. He intuitively understood. Yes, Google will still do its thing. But operational systems really can't drop ACID; even if Date suggests that such would be OK with him.

The days of centralized data, on a passive terminal device (VT-X00, for example), are back. I continue to be amazed how slowly the Well Known Pundits are coming to this realization. But they will because they must. It's not nice to fool Mother Nature.

No comments: