🚀 go-pugleaf

In article <1992Apr1.000442.748@debet.nhh.no>, si0_tb90020@debet.nhh.no writes:
> In article <25MAR199212413765@lims05.lerc.nasa.gov>, scdorcy@lims05.lerc.nasa.gov (JAMES DORCEY) writes:
>> 
>> Interesting.  Buy 3.1 and your recent 3.0 apps, if not made by Microsoft,
>> won't work, so you better buy Excel 4.0 and Word 2.0.  And make sure you're 
>> running MS-DOS 5.0.
>> 
>> Is there a pattern here?
> 
> Well don't bet your Microsoft applications will run. Today I tried to
> start an MessDoss Window from 3.1. Doubleclik .... wait .... coming up
> ... hanging.... ctrl+alt+del.... No response. The big click does work
> still though, and we are counting SIM chips...

I suspect that there is some problem (hardware or software) with your
machine.  I doubt _very_ seriously that there are significant problems
starting DOS windows under 3.1 -- it's been tested pretty thoroughly
by thousands of beta testers for something quite that basic to be _that_
thoroughly broken.  (Obscure incompatibilities after the DOS window is
started up are more likely to be overlooked, but the act of initiating
a DOS window itself is just too basic for such a bug to be missed).

One of the machines around here had numerous problems with 3.1 but had
not had with 3.0 -- this was eventually traced to bad chips (NOT memory
chips as you might expect but PAL CPU support chips).  Apparently 3.1
happened to exercise the faulty chips a bit more strenuously.  It's not
at all unusual for system software to be able to "find" hardware problems
better than the diagnostics - this is particularly legendary in the DEC
(PDP-11 and VAX) world, and I suspect will become more so in the PC
world as system software becomes more complex.

(BTW:  Whatever you think of the "Windows as OS" debate there is little
doubt that it is system software of some flavor.  Personally I think
that the two sides are arguing over the definitions of words -- in this
case the definition of "Operating System" -- and that when this happens
the discussion has become a politico/religious discussion analogous to
discussing how many angels can dance on the head of a pin.  What does it
matter to a user of the product whether it's an "Operating System" or
not?  Answer:  Not at all, what matters is how well it solves the user's
problems.  UNLESS you are involved in a religious war, then this can be
used as ammunition one way or another to obtain converts.  Please take
all such religious questions to talk.religion.misc or file 13, whichever
you prefer, but please leave the technical groups OUT of it).

It's also quite possible that there is some problem with your software
installation -- such as a HIMEM.SYS from DOS 5 rather than from Win 3.1,
or a TSR that does something flakey, etc.  I'd try dropping back to the
simplest possible installation, if that fails then your problem is almost
certainly hardware.  If that's the case then probably other complex
software (such as OS/2 or Unix) will also eventually have problems with
it & you should look into finding & fixing the problem. 

> Anyone said UAE's won't happen anymore? Sure thing, Microsoft just
> removed the errormessage, not the errors...

It's true that the UAE error message has been changed (to something
more explanatory) under 3.1, but it's not reasonable to expect that
ALL such errors will go away, EVER, under ANY operating system.  The
problem is that MOST of these messages, even under 3.0, happen because
of some application program that has done something illegal.  The problem
with the way that 3.0 handles this is NOT that it displays an error
message, but that it doesn't recover very gracefully from the illegal
action (an understatement - often it doesn't recover at all).  3.1 does
a lot more parameter validation and error checking, and can recover from
more sorts of attempts to write into random memory, etc. than 3.0 can.
This means that the error message under 3.1 is likely to be closer to the
point of the actual error (rather than simply reporting a problem caused
by corruption from an error that occured long before) and that Windows
can recover from the situation better than it can under 3.0.

3.1 is greatly more solid in its handling of application errors (read:
UAE's) than is 3.0, and it handles resources better (less chance of
resources filling up), but it does NOT do a good job of garbage collection,
so badly-behaved apps can still allocate resources that they never free
("leaking" resources), even when the app is terminated [!].  The situation
is much better than 3.0 but still pretty far from being ideal.

FWIW a number of windowing systems have the problem with leaking resources.
Part of the problem with this from the system designer's point of view is
that it's often possible for a resource to be shared by more than one app,
so garbage collection gets difficult (you don't want to delete the resource
until all references are deleted, but it's hard to know that unless you are
running a purely interpretive language).  It's still an important enough
problem that I think more attention should be given to it so that garbage
collection can proceed in at least a _few_ more cases.

Anyhow I've rambled on long enough,

Bruce C. Wright