[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: On linux kernel packaging issue, assuming that user is never right

On Sat, Nov 08, 2003 at 06:59:39PM +0100, Mathieu Roy wrote:
> Andrew Suffield <asuffield@debian.org> a tapoté :
> > On Sat, Nov 08, 2003 at 01:58:29PM +0100, Mateusz Papiernik wrote:
> >> Andrew Suffield wrote:
> >> >We're all very interested in *real* evidence here, because there
> >> >hasn't been any in the past. If you don't have any evidence, you can
> >> >expect people to call bullshit on this.
> >> 
> >> I can't send any *evidence* here, but I can post my own opinions and 
> >> experiences with kernels. And I'd say thats not a bullshit. Yes, that's 
> >> true that performance gain isn't very big and noticeable, but after 
> >> recompiling 2.4.18-bf24 from woody with my own optimisations for athlon, 
> >> I noticed little speed up in compiling my programs.
> >
> > This is probably what is known as "the placebo effect".
> >
> > Human impressions of this form are always entirely disconnected with
> > reality; the mind applies filters based on expectations, that throw
> > them hopelessly adrift. If you build your kernel with options you
> > expect to work faster, it will seem faster; if you take a pill, you
> > feel better.
> Why do you always assume being facing idiots?

Partial straw man compounded with slander. Learn some manners.

> People knows all about placebo effect, but do you have any evidence
> that there is nothing more than placebo effect?

I have considerable evidence that there *is* one here, and it's
Mateusz's statement, which I quoted

"I noticed little speed up in compiling my programs"

That's not "I timed it and it was faster", and it's not "I carefully
benchmarked it" - it's an impression. Whether or not there was an
actual performance increase is irrelevant, because there will have
been a placebo effect, and it renders the data point useless when
alone, statistically.

[You *could* build a valid experiment using "seems faster", but not
from one person and not in an uncontrolled environment]

> If you compile a kernel without lot of modules for hardware and stuff
> you do not have, there is nothing weird to suppose it may have some
> consequences at later point, by having a kernel size reduced by 25%. 

See, that's the sort of nonsense that I object to. It is highly likely
that kernel size does not affect system performance in a measurable
fashion, other than on low-memory systems. Why would it?

[The rest of your mail managed to miss the point entirely, so I'll
skip replying to it]

  .''`.  ** Debian GNU/Linux ** | Andrew Suffield
 : :' :  http://www.debian.org/ |
 `. `'                          |
   `-             -><-          |

Attachment: signature.asc
Description: Digital signature

Reply to: