[geeks] On the subject of mainframes...

[gsm at mendelson.com]

On Sun, May 29, 2011 at 08:30:58AM -0400, Lionel Peterson wrote:
>Gear comments, just a quick note that there was an ISPF/SPF alternative that
>ran on MVS, it relied on CICS and emulated ISPF/SPF - provided a very similar
>interface with much lower overhead & cost.

Was CICS ever free? I remember running CICS under DOS and then CICS/VS
and allthough CICS was open source (I actually had a copy on "my" system
and made mods to DFHTCP), I thought it cost money. 

>As for a reason to go back? Aside from nostalgia, I agree there were precious
>few programs that couldn't be run dramatically more efficiently under more
>modern hardware/software...
>
>A previous commenter wrote something to the effect of "I don't see how work
>was done back then" - it was easy, once you got past the idea of working
>interactively with a user. Writing basic mainframe batch code was like writing
>a really, really big UNIX utility, where each program ran on either flat file
>or database input, and generated either an updated text file(s) or database.
>Using JCL 'glue' and utilities (again, similar in some ways to UNIX shell
>tools) you'd string a bunch of single-purpose programs together to accomplish
>a given task. Also, jobs could be scheduled for execution via JCL by working
>with initiators (execution threads) that jobs could be released into, and then
>the queued-up jobs would execute in a first-in, first-out manner.
>
>Then there were interactive programs (CICS, TSO, IMS/DC were the big hitters),
>writing them was different from batch programming, but not unlike web
>programming now. 

CICS and IMS/DC were more like HTTP servers. They were transaction processors.
I once ended up in line for the bus behind someone discussing a patent 
application for a method of determining response time of web servers using
server resource measurement. I mentioned that we were doing it in the 1970's
with CICS and other mainframe systems. It was not a fun ride home.

TSO was a real timesharing system, you could enter commands, have them
execute and they were things you would do. There was no email or netnews,
because email had yet to be invented. We used to send messages by putting
text in commonly readable files. (circa 1969)


The program would run, throw up a screen, and terminate. The
>user enters data, hits an action key, and based on the key hit, the same or
>another program would run, process the screen contents, update databases, and
>then draw a new screen and end...

Yes. 


>Mainframe tools were wonderful - and facilities like PDS (partitioned data
>sets)
PDS are more like ar archives under unix, but were used like sub directories
as such things did not exist.

VSAM files,  VSAM was a simple indexed file system. There used to be C
libraries to do the same thing, but I have not seen them this century
B-trieve?

IMS/DB was a database engine, hirearchical instead of relational. Look it up
if you care. :-)



  were great, and the JCL utilities were very,
>very powerful, enabling one to accomplish a lot of work without writing any
>code...

jcl was basically a big shell script, as you put it. I think both TSO and
CMS had some sort of scripting language, and the pay-for-it CMS (VM system
product) had REXX a great interpreted language.


>
>Ah, memories...
>
>Lionel
>
>On May 28, 2011, at 5:47 PM, gsm at mendelson.com wrote:
>
>> I don't see any thing that a modern person would be interested in.

I still don't. I remember all of that stuff fondly, but there is a reason
why modern tools evolved over the last 40 years. 

Geoff.

-- 
Making your enemy reliant on software you support is the best revenge.