From: meshkin@sol.cs.wmich.edu (Parental Advisory) Subject: Re: NT vs Linux (was: Re: truth or dare) Date: Fri, 9 Jul 1993 08:10:37 GMT
On 8 Jul 1993 18:15:19 GMT, Jerry Shekhel (jerry@msi.com) seemed to say:
> Howlin' Bob (gt8134b@prism.gatech.EDU) wrote:
> : rgallen@muug.mb.ca (Rennie Allen) writes:
> :
> : >There is more to quality software than code that will pass through lint
> : >without a burp. There is *design*. Linux has no *design* it is simply a
> : >clone of what has gone before, it does nothing to advance the state of the
> : >art in OS design.
> :
> Well, maybe it's time to take a break from trying to advance the state
> of the art. Maybe instead it's time to try to come up with something truly
> useful -- something that pleases its users at least as much as it pleases
Hooray! My sentiments exactly.
One of the things that has irritated me for quite some time is all the
theory people telling us how we have to make our code more "generic",
"reusable" and "modular". While this is very nice for general user
code that might be ported to a dozen different machines, or for writing
things that are similar and can share routines, it can also get out of
hand very quickly and result in bloated slug-code. I trust you don't
really WANT your kernal to be generic and slow?
Does anyone out there remember the kinds of things that used to be
accomplished back in the days of the Commodore 64, Apple ][, etc?
Remember when we only had 64K in the whole machine? Remember when
a processor ran at a whopping 1MHz? Anyone remember 88K floppies?
It seems to me that many people write very sloppy code and tell us we
need to buy faster hardware to make it work. Being a CS person myself,
I quite understand how the underlying algorithm is important. Certainly
something based on a heap-sort CAN be much faster than a bubble-sort...
but I can also write a super-inefficient implementation of that heap
sort which WILL fail to beat a nice quick assembly bubble sort.
We need to use state-of-the-art techniques, but we also need to use the
best optimizer there is, our brains. From what I have seen so far, Linux
is doing this far better than most of the commercial vendors.
Just my 2 cents tossed into the flames...
-Chris.
--
____ ____ ____ To seek out new structs, and polymorphism...
// \\ // \\ ||// To boldy go where no to has gone before...
\\__// \\__// ||
Chris Meshkin (meshkin@sol.cs.wmich.edu) > WMU <