[vox-tech] Three Install Questions

[Karsten M. Self]

on Mon, Feb 21, 2005 at 12:10:28PM -0800, Rick Moen (rick at linuxmafia.com) wrote:
> Quoting Peter Jay Salzman (p at dirac.org):
> 
> > Other than a story of a poorly understood case where multiple swaps per disk
> > seemed to suppress the expression of a VM bug, can you describe what benefit
> > multiple swaps per disk give?
> 
> Nope.  
> 
> But that reminds me of a point that might be tangentially relevant:  In
> trying to help people with Linux problems, one tends to find that people
> get themselves into the _damnedest_ fixes, i.e., peculiar or brand-new
> or poorly supported chipsets, bizarre system configurations, and in
> general problems that make you scratch your head and wonder "Why on
> _earth_ would someone be doing that?"
> 
> I mulled over that for a long while, and have built up a hypothesis: 
> Those of us who've been doing this for a long time tend to, without
> really thinking about it, semi-automatically avoid most of the
> aforementioned troublesome situations through habits driven into our
> skulls by experience.  This doesn't make us smarter or wiser, but it
> does make us, in general, _luckier_ within a limited sphere.

It's a lesson that's been burned, drilled, pressed, embossed, and
tattoed on my own skull a few times.

ObDigression:  in one of his biographies (IIRC:  _Surely You Must Be
Joking_), Richard Feynman describes the process of designing some
electromechanical device.  He's presented with a gear chart.  On asking
for guidance as to which gears to use, he's told:  Pick the ones from
the middle of the chart:

  - The smallest gears have too few teeth, and wear out too quickly.
  - The largest gears have small teeth, which break too easily.

The stuff in the middle, however, is well within the design envelope,
and tends to work.

Similarly, there's a lot made of extremes of design:  biggest aircraft,
tallest skyscrapers, smallest microdrives, etc.  The problem is that
such creations exist on the edge of the design envelope.  Frequently,
even where practical from an engineering standpoint, they're not
economically feasible.  Viz:  

 - The World Trade Center Towers, NYC.  This was actually a government
   project (NY/NJ Port Authority) and was not commercially successful.

 - The Space Shuttle.  Impressive engineering accomplishment.  Great PR.
   A hell of an expensive way to lob a metric tonne into orbit.  And
   when failure does occur (which it inevitably does), you need another
   seven astronauts.  BTW:  you've got a 3% probability of death on a
   given launch, based on 40+ years of data.

 - Various large aircraft.  Spruce Goose.  The new Airbus A340 will be
   interesting to track.  Boeing's taking the track of smaller aircraft
   with longer flight ranges.

 - Nanotech.  I first started hearing about this in college, some 18
   years ago.  My thought then matches my thoughts now:  interesting
   stuff.  But we're talking about design at the physical limits of
   structural integrity.  Manufacturing low numbers of devices is not
   likely to be practical.  Look at Neal Stephenson's _The Diamond Age_
   for a reasonably plausible scenario.  And current applications tend
   to be more in applied material areas than devices, per se.  Textile
   coatings are one of the biggest successes -- wrinkle-free fabric.


I've also had a few experiences with GNU/Linux that have generally
pointed me back to x86 hardware based on various forrays.  Sure, it's a
klugy design with a lot of faults, but everyone knows those faults, and
builds stuff with it.  Next-gen hardware will emerge, eventually, but I
predict it's going to be x86 compatible for years.

 
> Thus:  It would never have dawned on me to create a single 32GB swap
> partition.  I'd have instead, left to my own preferences, made three
> or four smaller swap partitions if for some reason I needed even half
> or a quarter that much total swap.  I'd not have been able to tell you
> exactly why, but, if I sat down and tried to puzzle out why for a long
> time, it would have been something like this:  "Well, I figure that
> 256MB to 1GB swap partitions have been reasonably common for a while,
> as have multiple swaps per system, but single gigantic ones probably
> haven't, which means the latter configuration might end up using code
> in ways that aren't broadly tested, which means that it might trigger
> fault conditions not yet debugged because they've been seldom seen."
> 
> And, in that _particular_ situation, with that particular kernel VM, my
> instincts would have been right.

Oi.


Peace.

-- 
Karsten M. Self <kmself at ix.netcom.com>        http://kmself.home.netcom.com/
 What Part of "Gestalt" don't you understand?
    This suggests a more troubled future for Gnome than I had imagined.
    - Chip Salzenberg, on GNOME developer attitudes
      http://zgp.org/pipermail/linux-elitists/2004-January/008588.html
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 189 bytes
Desc: Digital signature
Url : http://ns1.livepenguin.com/pipermail/vox-tech/attachments/20050222/2b705846/attachment.bin