Wireless Strength: Windows vs Linux

[jc at trillian.mit.edu (John Chambers)]

Don Levey commented:
| I'm noticing something odd on my wife'a laptop.  I put a DLink wireless card
| in it so we can move it around the house; in Windows XP the reported signal
| strength (per the OS itself) is always very strong.  However, in Linux
| (Fedora 4), the reported strength is mediocre at best.  The antenna is
| perhaps as much as 20-30 feet from the laptop, and even moving the thing
| into the same root 3' away brings me only to about 50% signal strength when
| in Linux.  Is this just a function of the reporting tools I'm using
| (wlassistant, KDE panel applet), or is there truly some sort of signal
| enhancement going on in Windows?

I've seen a number  of  signal-strength  indicators,  and  I've  been
impressed  by their poor quality.  In particular, I've yet to see one
that actually has any sort of unit attached to the strength.

At home, I have a Mac laptop that has two such signal indicators, one
called  "MacStumbler"  ("Stumbler",  a  clone  of  NetStumbler that's
available for linux/BSD) and Internet Connect ("IC").  Stumbler shows
a number; IC shows a line of 15 "bars" that are two different colors.
Wandering around the house, I find  the  following  signal  strengths
claimed by these two programs:

        Stumbler  IC ratio
Place 1       39  15   2.6
Place 2       29  12   2.4
Place 3       25  11   2.3
Place 4       32  10   3.2
Place 5       30  10   3.0
Place 6       30  12   2.5
Place 7       36  10   3.6

As you can see, they don't agree very well at all.  The  numbers  are
not  just non-proportional; they move in different directions.  I did
let them sit for a while at each place,  until  they  stabilized,  so
it's not that one is still measuring when the other is done.  They're
just not in agreement about the signal strength.

One inconsistency isn't visible  in  those  numbers:  Stumbler  shows
numbers as high as 80 when close to the hub, while IC's max is 15. So
the first line of numbers above has Stumbler showing roughly  50%  of
max,  while  IC's  number  is  at  the  max.   This  doesn't  inspire
confidence.

It reminds me of the old saying: If you have one watch, you know what
time  it  is;  if you have two watches, you're never sure.  The above
inconsistencies make me seriously doubt both of them. If either had a
decent explanation of what the numbers mean and how they are derived,
maybe I'd trust it a bit more.  But they don't, to my knowledge.

Anyway, I'm not impressed by these, or any other similar  tools  that
I've seen on various wireless machines.

(I should get a wireless linux laptop, so I can criticise  its  tools
similarly.  ;-)