anton (06/20/82)
This is reply to houxf!kenig's flame about me not liking TSO
I like the subtle poke at UNIX and its uc/lc "nonsense". Hey Guy, its not
UNIX, its ME ! Just like I often feel to lazy to use nroff to neatly
format my mail, or the speller to check it, I sometime forget (read am too
lazy to bother to..) to use the shift key. At least I have a system
that can differentiate between the case of commands !
The name is *anton*, ***NOT*** Anton. I am on a humility jag at the moment.
I fully agree; UNIX is just as much bullcrap as any other system. However
it has an important difference - UNIX is useable.
To assert TSo is useable is not sacrelidge, it is sheer stupidity. My last
contract was developing 8085 code on a TSO link to a pair of 3033's.
Never in my life have I used a system with such a poor console response,
such a poor set of development tools, and such a "Not Invented Here"
attitude - the same one you accuse me of ! A simple task that should have
taken 6 weeks spread out to nearly six months. The costs involved in the
link over that time would have more than paid for an Intel MDS which would
have been better equiped than that service. I like UNIX because I can do
so many non trivial things with it, like write compile test and document
my code - which many IBM instalations seem to think isa funny thing to
want to do with a computer. I can get a response within less than 10 seconds
even when the system is heavily loaded, which is mush better than TSO.
My current record is 23 minuites for a search command while editing an 8K file.
Specific gripes:
I had the FORTRAN output of the RATFOR version of the UNIX ed. I
could have got it loaded, but I could not have used it under TSO.
You have to specify the files in the JCL, which is outside the
programme. For much the same reason, I couldn't use the
'include' directive in the 8085 assembler.
After some editing, a whole new copy of the file is made, without
releasing the space used by the original in the partitioned data set.
This 'original' is not for 'back-up' purposes, & I could never get
at it. Often, at the end of an edit session, I would do a save,
and the system would tell me there is no more room in my dataset.
Yes, I could compact it, but it was an involved process. Demand
allocation and freeing of file space ? Not Invented Here !
(I'm not going on with this, its making me feel angry !)
I am not a diehard UNIX-phreak. I have used many OS's, in particular the
very experimental & customised ones out of Cambridge and Canterbury, mostly
written in BCPL and offering a higher throughput on an old English Electric
machine with 3 microsecond store than I get of these souped up
microprocessors with 0.3 microsecond strore - and a lot more of it so that
they don't have to page. I know of a lot of things I would like to redesign
in UNIX, since suporting UNIX is my full-time (!) job and I would like to
make it a LOT easier. However, there is company policy to deal with, which
is why we run UNIX & witch co-processes rather than TUNIS and use the much
easier to understand and maintain monitor structures. TUNIS can emulate
UNIX, and more efficiently as well.
"As a science we know less about operating systems
and what makes them effective tools than we do
about programming languages and what makes for
a good programming language."
So what ? We do, however, know a lot about what makes for a convenient system
to use, what tools and methods are actually useful, rather than just for
glossy brochures - viz Visicalc and other screen oriented methods.
viz the work Tony van Dam is doing at Brown on a visually oriented shell.
viz "The Blit" from Robert Pike and Murray Hill.
Damn it man, viz UNIX against all that was extant when it came along !
An open mind ? The trouble with that is that people will come and drop all
manner of garbage in it, starting by inflating your self esteem and saying
that this magic potion will help even more. Hence my humility jag so they
can't catch me. Like Lazarus Long said, there are many advantages in being
lazy, if you are intellegent enough about it. I am in a well paid job, in
a city where I can walk the streets at any hour of the night without fear
of muggers, where food and entertainment are cheap, the girls are pretty
and my OS is reasonably intellegent.
"Most of the B.S. you pander about as professional
style (be it from UNIX or anywhere) simply strikes
me as goddamned common sense. Who would want to
incur the wrath of the people you see every dey,
be they fellow employees or customers by not
providing documentation with the code? "
Sadly, I have worked in too many places where the attitude to both hardware
and software was 'get it out the door so we can invoice the customer'.
This is the price you pay for having firms run by accountants and
borrowing from banks. On top of that, there are establishments where
the only way to ensure tenure is to make sure ONLY YOU can fix the code,
and there are bugs there for you to fix.
I don't hate IBM any more than I hate the moons of Saturn, but the environment
that IBM enbcourages, both in house by its management and sales and 'internal
rules on coding and documenting' are BAD for the industry. Which, when you
look at some of the neat tricks of hardware, invented in the 40's and 50'
at Manchester but totally impractical they have turned into working and
profitable apporaches to computing, is a crying shame.
example: RULE: MODULARISE
IBM Version: modules must not use more than 4k of memory
(hmm 360 offsetaddress range ??)
action: hacker makes sure his code only fits in 4K, and it does;
however, it is actualy 28K of code, and calls the swapper
endlessly, so clogging up the system.
(See Brooks "Mythical man month" for details)
UNIX is good because it doesn't tie you down or make unwarrented assumptions.
As soon as you get big you have to have a lot of organisation, and the
entropy of the system is that this aspect predominates. (e.g the great 60's
idea of "management by objectives" eventually clogged because it ONLY managed
to produce... yes, you guessed it. 'management by objectives'. THAT is why
Townsend ("Up The Organisation") was so sucessful.
See also:
"Entopy: a new world view" (I forget the author)
"System-antics" (John Gall)
"The Peter Principle" (Dr. Lawrence Peter)
To conclude: I know we can do better, even with what we have at the moment,
but economics means we cant throw out all the braindamaged
pdp-8's and naked minis and all the rest, never mind al the
Gigabucks worth of programmes in FORTRAN &COBOL which would
be more efficient and more maintainable in PASCAL, ALGOL
or <insert your fave>. Add to that the fact that most
managements are well entrenched and utterly opposed (for
a variety of reasons, but mainly loosing face (which is why
I am on a humility jag) or thinking they have over a past
decision that now needs to be retracted, and you get the
state of the world as it is. Tough, but we have to live with
it. Or do we ? The pressure from UNIX's popularity
spawned the "Virtual Operating System", which could
revolutionise TSO, CP/M,....... and force manufacturers
to look to users actual needs rather than their hypothesised
ones.
maybe sometime soon I will, as I know I can, design a
no-compromise machine: banks of 16032's 32Meg physical,
bubble memory instead of disk. Its all feasable, and not
having to cut corners to make it marketable, and not having
to keep to anyone's standards for the UNIX kernel or anything
else, I think I can come up with a supermachine that costs
less than $10k. Hmm. I wonder if in the long run I can be
lazier if I have better automated programme generation from
specifications using PSL-PSA, and EUCLID with a "prover" ?
/anton ..!decvax!hcr!anton