25 Apr
2012
25 Apr
'12
2:22 p.m.
Now this is getting silly. There is no point in
trying to say that what
Tony is seeing is wrong: unless he is psychotic he is actually observing
I am very happy to accept that human observatiosn, even when not
deliberately mis-represented can be unreliable [1].
However, I regard it is very foolish to discount things that I have
observed (or believe I haev observed' based on 'This is the real version
becuase I am telling you so' type messages. I have even less evidence
that those are correct.
something that really has happened to him. And all
that "I have this
fancy job in industry and I have signed lots of NDAs, so I am important,
while Tony is just a geek living with his parents" is a) rude and
I haev long since ignored the insults here. I simmple remember a proverb
we have in English. 'Empty vewssels make the msot noise'.
patronising and b) hiding behind these NDAs you claim
to have signed. So
go out and find publically available research reports or whatever on the
Internet to prove your point, instead of trying to impress us with smoke
and mirrors.
Exactly. I accept that reliability data is commercially sensitive, and of
course, fi you agree to do something (or in this case not to do something
-- not to reveal that data), then of course you must honour that. But
without seeing the data, what it applies to, how it was obtained, how it
was processed, etc, I ahve no reason whatsoever to accept anything based
on it.
Of course my sample is biased. But in a way, it's biased towards the sort
of things I work on. I don't get to see infant mortality or production
failure rates _but_ those don't affect me when I am repairing production
machiens in the field. And for that matter , Iwould argue anybody's
sample is biased for similar reasons. It applies to particular conditions
(for example, failures in the pre-shipment testing). And without knowing
what it applies to, the data is meaningless.
However I think <unknown> is probably right in one respect about modern
computers being more reliable: considering the complexity of a modern
$500 PC, it is probably much more reliable than anything made in the
'70s *at a corresponding price for the period*, i e today's cheap junk
is more reliable thatn cheap junk from the '70s and '80s. OTOH Tony is
also right that old computers were/are at least as reliable: they were
better made than today's cheap junk.
There are 2 points here that I will comment on...
Firstly /consideerign the complexity'. I think I stated very ewarly on
that a VLSI IC is more reliable than making the same circuit out of MSI
ICs (assuming you could) which, in turn, is more reliable than making the
same circuit out of discretes. I doubt anyone disagrees with that. The
point is 'same circuit'. That does not mean the same function. To give a
silly example, you can flash an (LED + current limiting resistor) using a
microcontroller. You can also do it with a 555 2 R';s and a C. Or with 2
transisitors, 3 more R's and 2 C's. It is not inheerrently obvious to me
which of thos is goign to be the most relaibvle, it may depend on the
sort of reliability you are considiering ('how many work driectly after
production' .vs. h'ow many work after a year' .vs. 'how many are stil
lworking after 25 years').
The second point is 'for the price'. PCs are very cheap for what htey do,
I will agree with that. And to compare them against the DECs, HPs, etc of
twety, thirty or forty years ago is unfair. However, the fact that a fair
numbero f those machiens are still running, while there are capacitor
provlems (say) on modern PC motehrboards does say somethign for the
reliability of the older machines...
-tony