Re: Market pressures for more reliable software

["Shmuel (Seymour J.) Metz" <nospam@acm.org>]

Lao Xiao Hai wrote:

> I must be missing the point.    What was not centralized?

As I said multiple times, programming.


> Programming was pretty
> much centralized and under the control of a data processing manager.  Computers
> were pretty much centralized, and there weren't that many that could drive remote
> terminals really well.

Well enough. Can you say RAX? QUICKTRAN? CRBE? CRJE? RJE? SGJP? ATS? HASP Multileaving? And
that's just IBM supplied. There were also ALPHA, ROSCOE (nee WRAP) and Wylbur on S/360, and
others on non-IBM hardware. There was enough time sharing and remote batch to drive a market
in plug-compatible terminals and modems.

> Most of the data processing was in batch mode.

Which doesn't make it centralized.


> That is not how I remember it.

Clearly. Either your experience was limited or your memory is faulty.



> Most of the programs I recall were written by the
> programmers
> at the service bureau.  Some were packages purchased from elsewhere.   It was the rare
> customer who had programming resources on which they could rely for their own software.

By the 70s most service bureaus were offering remote access. Ever wonder why?


> Not that common.

Common enough to drive a market in brand X clones of terminals and modems, e.g., Cope 45 for
remote batch, Vadic modems.


> And it was expensive.

TI 700s were dirt cheap in the 70s and 80s, as were acoustic couplers.


> I have done both.   Certainly a lot of scientific computing was done using remote
> computers.   However, I recall quite vividly those sites that had their own IBM 1130
> and did their own Fortran programming.   I spent many late night hours writing Fortran II
> and debugging on the 1130.

And how many of those 1130s cost $150,000? To say nothing of the fact that a lot of them were
submitting jobs to larger machines.


> The IBM 1401 was the workhorse of industry for a long time.

And cost nowhere near $150,000.


> Oh, and you forgot one of my personal favorites, the CDC
> 160 series.

I didn't forget it, any more than I forgot LGP or RCA; I omitted it because I was
concentrating on the IBM marketplace.

> The IBM System 360 series came along in 1964 and, along with its
> clones and cousins, did required lots of memory, lots of dollars, and lots of
> programming talent.

The S/360 may have required lots of memory, but the most it would take was 64K. The 360/20 was
even smaller.


> I am noticing that some very large organizations are moving back to a more centralized
> computing authority.    User developed applications  (UDA's) are getting out-of-hand
> and instead of saving money, are costing money.     Worse, many of the amateur
> programmers who create UDA's are failing to build in the kinds of error control and
> security one would expect of more rigorously defined software.   This is creating
> vulnerabilities that would not have been tolerated in the centralized era.

Those shops that were centralized in the old days had their own horror stories. Things like
configuration management and backups were often slipshod, if they existed at all. Some of the
decentralized shops had their act together better than some of the centralized shops.

--
Shmuel (Seymour J.) Metz