Re: Is there a mainframe skills shortage?


<docdwarf@xxxxxxxxx> wrote in message news:euil03$l33$1@xxxxxxxxxxxxxxxxxxxx
In article <57147bF29lbfkU1@xxxxxxxxxxxxxxxxxx>,
Pete Dashwood <dashwood@xxxxxxxxxxxxxxxxxxxxxxxxx> wrote:

[snip]

I don't think so. It is much better today than it was thirty years ago. I
can remember getting bright eyed, bushy tailed, Computer Science grads,
sitting them down and asking them to do a three file merge in COBOL...

That's interesting, Mr Dashwood... my first response to that was 'there's
no need to do it in COBOL, one uses system utilities (DFSORT/SyncSort) for
such tasks... but if one needs Lines O' Code to satisfy a Corner Office
Idiot then there's a single verb to manage it'.

Be that as it may... I was taught that Computer Science grads should be
designing compilers while applications jockies write the COBOL.


Sure, and guys go out to work while women mind the kids and cook dinner. A
somewhat arbitrary and constraining division of labour which, like not
letting CS grads develop applications, may prove less than optimum.


Most of them didn't even know how to approach the problem, yet these same
guys could write a random number generator based on a binary polynomial
that
would guarantee never to repeat within so many billion samples... (Not a
lot
of call for that in the Banking industry :-))

That's why one doesn't hire CS grads to write banking apps... seems like
our educations were different.

You conveniently ignored my use of the past tense... the whole point of this
particular discussion was that grads today are better than they used to be,
at least that has been my experience.

Leaving that aside, I have no doubt that our educations were different :-)



Today, they would look at you and say: "Why would you want to merge these
sequential files when you could have updated directly to a Relational
Database in ANY sequence you like and simply ORDERed the result set?" :-)

'Because two of the files are the inputs from branches with a lot of
activity, (n) tens of millions of records per day; they are MERGEd into a
single dump of the master table and reloaded, directly, into a re-creation
of the master using a system utility. Do you have any idea why that might
be superior to performing individual INSERTs?'

Actually, it wouldn't be, using todays technology. The
inserts/updates/deletes are done on in-memory objects that are part of a
result set, the master doesn't need to be re-created using utilities or
application code, and the number of "records" per day or per second is
pretty irrelevant. The new technology and SQL is geared to parallell
processors... you just add as many as you need to handle your load. But
today's CS grad would understand all of this stuff and find no problem in
using Lambda expressions to manipulate in-memory sets.

It is indeed a wondrous and changing world we live in.

Pete.


.