Subject: Re: [executables] was: why Haskell hasn't replaced CL yet?
From: Erik Naggum <erik@naggum.no>
Date: 2000/03/03
Newsgroups: comp.lang.lisp
Message-ID: <3161062467384796@naggum.no>

* Samuel A. Falvo II
| With the computer doing absolutely nothing else in the process.  Do you
| see the difference here?  In a real-world situation, a computer could be
| handling network connections, performing back-ups, or otherwise running
| other CPU-intensive tasks in the background.  That all will definately
| influence (and lengthen) startup times.

  oh, geez, when will this _end_?

  I don't have this powerful a machine just to boast about it and play
  games.  it's a work-horse for serious development, and it has a number of
  non-trivial duties.  at the time I ran those tests, it turns out that it
  was servicing a few thousand FTP requests from local network machines
  that were upgrading some software automatically over the span of the few
  minutes I ran the tests, it ran a bunch of Netscape frames with animated
  advertising GIFs, and it provided monitoring and backup services for 6
  other computers on its local network, which involves network traffic and
  low CPU consistency checking.  it also received four e-mail messages, the
  processing of which fires up an Emacs in batch mode to handle the
  filtering and processing of the incoming messages.  the only thing not
  strictly normal about this is the FTP load.  regardless, I have no idea
  exactly how big this load was during each of the individual _seconds_
  that I ran my tests.  I have reason to suspect that it had very little
  effect on anything because the machine is in fact able to perform the
  vast majority of its duties in zero noticeable time -- which is why it is
  this powerful to begin with.  now, this _could_ explain the 5 ms extra
  execution time I noticed, but that's just pure speculation on my part,
  and I see little point in spending the time to figure it out.

| I was just pointing out that the measurements performed could be
| misleading due to the circumstances in which the measurements were made.

  so let's assume the measurement errors were on the order of 20 vs 25 ms
  per invocation.  that's the difference between 40 and 50 invocations per
  second.  this bothers you a great deal, apparently.  it doesn't bother me.

  and you were "just pointing out" that it could take _minutes_, which is
  nothing more than really bad fiction on your part.  in _minutes_, this
  machine has compiled GNU Emacs from scratch (2:30), built a new Linux
  kernel (2:10), installed staroffice (1:20), built CD images for Debian
  2.2 (3:10), or upgraded and installed a 100 packages (2:50).

  to suggest that this machine should suddenly only manage to start 50
  Allegro CL processes because of other work it's doing is simply insane.

  as long as any goddamn fool can cast doubt on anything anybody says, I
  suggest a much more honest starting point: "I don't want to believe you!"
  instead of trying to smear whoever is trying to answer their questions.
  I'm getting sick of the rampant stupidity that comes with benchmarks and
  any other myth-deflating devices.  myths, apparently, are necessary for
  the mental survival of some people.  perhaps it is not a good idea to try
  to destroy their misguided beliefs because they turn out to be lunatics
  if they can't keep their myths alive and well.
  
#:Erik