Warp wrote:
>   OTOH, 20 years ago most serious operating systems (and consequently other
> software) were running huge servers with hundreds, if not thousands of
> users.

... but still running a fairly small collection of software compared to 
what tends to run on machines today. The Linux boot partition today is 
multiple times the size of the biggest hard drives you could put on 
"desktop" computers 20 years ago, and orders of magnitude bigger than 
the disk space available on a mainframe 40 years ago.  10 years ago I 
was impressed when the boss brought in a Windows machine with 128Meg in 
it. Nowadays you can't even boot Windows in that, let alone do something 
useful.

And of course, the hardware is flakier too. That's why NASA puts 
20-year-old technology into satellites and space ships. You read the 
google papers, and you see they run some job and it takes on the average 
1000 machines, 7 minutes, and during that time on average two machines 
will have some sort of hardware failure. Your software can seem really 
flakey when your swap space has an unreliable sector in it - btdt.

Plus, software changes much faster today.  People in a university aren't 
going to use a compiler last updated (except for bug fixes) 10 years 
ago. Nobody these days uses a C++ compiler last updated before the STL 
was created. Nobody uses a database engine that has been stable for 10 
years, or even an OS that has been stable for 10 years.

It's easy to make a system that doesn't crash. Get it working, then stop 
dicking around with it. Windows servers are quite reliable, for example, 
because people don't install new stuff and then take it off, plug in 
random USB devices, try out new screen savers, etc. They're very careful 
with their upgrades, just like they are for Linux.

>   This was rather common especially at universities and other similar
> academies: You were lucky if you had access to an actual desktop computer.

Given the power of desktop computers 20 years ago, I'd disagree with the 
"lucky" part of that. ;-)  The stupidest thing the CIS department did at 
my undergrad school was to start teaching assembler on Commodore PETs 
instead of on the mainframe.

-- 
Darren New / San Diego, CA, USA (PST)

Post a reply to this message