2/11 Responce to C, C++, new grad, love for comp sci, love for $$$, 1 word:
\_ The smart ones learn to spell or run a spell checker.
\_ And how to count words.
The dumb and greedy ones work in the industry
\_ think C++, Bell Labs.
\_ C++ is a perfect example of industry stupidity.
The smart ones go back to academia
It is as simple as that.
\_ the smart, greedy ones go work in 'XYZ labs'
\_ Have a cookie, troll.
\_ This is not true, as all of the major contributions to cs have been
mostly performed by industry. Without AT&T there would be no Unix
(BSD or otherwise). Without Sun we would be stuck with RFS.
\_ And without Berkeley we wouldn't have the 50 million BSD
derivatives, RISC, RAID, IEEE 754, yadayadaya. Incidentally,
they all seem to come from Berkeley. So you can conclude that
all major contributions come from Berkeley and industry.
\_ Stanford is usually given equal credit with UCB
for RISC (look at Patterson & Hennssey for example)
\_ I would concur that the Berkeley <-> Silicon Valley has
created the Academic-Industrial Complex.
\_ RISC? RAID? 754? Hardly as important as UNIX. Nice but
trivially obvious and would have been done by someone, some
where in due time.
\_ Two other contributions of industry include OpenFirmware
and FireWire. I don't think that academics ever came up
with ideas like those. But at the same time, X windows
was quite a good idea from academia. Sun would have saddled
us with news or openlook or some other stupid interface.
\_ Uh, I think most people agree that the X architecture
is ... well broken.
\_ NeWS was a much better architecture, but X was open
source, so it won unfortunately.
\_ RIDE BIKE! wins again for no particular
reason. One day the 'best' software will
win. Not the most politically correct.
\_ Not sure I agree. When it came out it was amazingly
overweight and bloated, but by modern standards,
it's fairly svelt, and it's suited to transparent
network redisplay, which I thank X for almost
every day.
\_ Oh, and there was also that stupid visiting prof who
discovered how to matrix multiply in O(n^2.7) instead of
O(n^3) which in many implementations turns out to be
slower anyway but only wacked out math people like ilyas
care about crap like that.
\_ Variations on Strassen's algorithm are pushing on
O(n^2.3) or something like that now. At any rate, while
I am not sure if anyone actually uses Strassen's in
practice, I do know that sometimes a tighter upper bound
on the running time can make all the difference in the
world. Fast Fourier Transforms and Pearl's belief
propagation algorithms come to mind as good examples.
-- ilyas
\_ Strassen's is pointless on today's hardware,
where mults are as cheap as adds. In the past,
\_ aren't multiplies still slightly more expensive
due to the 32 or 64 element carry save adds.
i don't think any digital design allows for
a 64 level deep logic in one clock cycle
for power and performance reasons.
and maybe again in the future, Strassen's
might actually be useful :) -nick
\_ Yawn. Oh, were you saying something?
\_ Shove it, asshole. If you're so ignorant as to
completely disregard the importance of strict
mathematical thinking in CS, go to the industry,
bury yer ass in some QA dept and be a mindless
drone as much as you wish. Otherwise, have a
cookie. -- not ilyas
\_ tell us about the stars
\_ "Do it my way or you're an ignorant QA
drone!!! My way is the only way to the
Purity and the Truth! I am CS Tao!!!"
Oh uhm, were you saying something? I
was distracted by a passing fleck of
something more important than your
opinion. I think it was a bit of pocket
lint floating gently to the floor.
\_ Or Andrew. God forbid!
\_ or Mach and Kerberos.
\_ Do the smart ones go back to academia and GET A FUCKING DICTIONARY
or SPELLCHECKER?! --greedy dummy in industry with SPELLCHECKER |
http://csua.com/feed/