Thank you, orinoco_cs

Wed Sep 25 11:40:05 EST 2002

On Tue, Sep 24, 2002 at 04:15:19PM -0700, Jim Carter wrote:
> On Tue, 24 Sep 2002, Timothy Murphy wrote:
> > (1) I'm not clear on the requirements for mode, ESSID, etc.
> > In my case it does not seem to matter whether I set mode
> > to ad-hoc or managed;
> > I cannot set mode to master, or change the ESSID,
> > at least while the card is active
> > (as opposed to changing the relevant entry in /etc/pcmcia/wireless.opts).
> 
> The MAC address of 02:02:2D:4A:52:80, is that the address of either of your
> cards?  If not, it's the address of your neighbor's access point :-)  My
> /etc/manuf file does not list 02:02:2d:*, but 00:02:2d belongs to Avaya.

The 02 bit in the first byte of a MAC address is the "locallly
administered" bit.  In Ad-Hoc mode most cards seem to set the BSSID by
taking the MAC address of the master node and setting this bit.  So
that BSSID is expected.

> The mode should be Managed if you're using an access point, or Ad-Hoc if
> not. In Ad-Hoc mode, both (all) cards must have the same ESSID and channel.
> In Managed mode you don't set the channel (the AP setting is
> authoritative), and some cards will let you set an ESSID of "any" to
> connect to the first network encountered, and some cards have this setting
> as the power-on default.

However in some situations where I'd expect ESSID="any" to work it
hasn't, but setting an explicit ESSID has fixed things.  I suspect we
may have some bugs in this area, at least on some firmware revisions.

> > (2) What exactly is the significance of the entries
> > for Link Quality, Signal level and  Noise level ?
> > Are they accurate;
> > or if not, is there any other way of determining
> > how sensitive the card or cards are?
> 
> Signal and noise level are in dB relative to a non-obvious origin, so an
> increase of 3 dB (i.e. less negative) corresponds to doubling the signal
> (or noise).  "Link quality" means the difference of these, the signal to
> noise ratio, which is what really determines how fast the link can run.
> Over 10 dB is marginally adequate; over 20 dB is good.  In Managed mode the
> statistics for the access point are reported automatically.  In Ad-Hoc mode
> they're all zero unless you do "iwspy eth0 $adr" where $adr can be either
> the MAC or IP address of the station you want stats about.  In your case it
> looks like the cards can't identify the access point, and so are giving all
> zero.
> 
> The dB origin for signal or noise varies from one vendor to another, but if
> one brand of card reports a higher "link quality" for the same partner, it
> will lose fewer packets or can operate at a greater distance. If the
> partner's card is replaced, with the same receiving card, a higher number
> means the partner is transmitting more power. 33 and 50 milliwatts are the
> powers I've seen mentioned on the list. Check the tech specs of your card,
> in the printed material or on Avaya's web site.

Well, that's what they're supposed to be.  I don't have firmware
documentation to categorically pin down what the values returned by
the card mean.  So I strongly suspect we misinterpret them in many
situations.

> > (3) On a slightly different tack,
> > what is the relation between Avaya, Agere, Lucent, Orinoco,
> > Wavelan and IEEE?
> > Has Lucent licensed the Orinoco hardware or software
> > to companies such as Avaya?
> 
> In the USA, AT&T was forced to divest itself of the Regional Bell Operating
> Companies, but kept Bell Telephone Laboratories, Western Electric
> (manufacturing), and the long distance business. Later it spun off Lucent,
> which included Bell Labs and Western Electric. During the dot-com bubble,
> Lucent decided to concentrate on its core competencies and not have its
> balance sheet weighed down by dull, plodding, old-tech manufacturing
> organizations, so it spun off Avaya (Western Electric: central office
> switches, cabling systems, connectors, etc.) and Agere (microwave
> manufacturing). In its heyday Lucent traded for US$ 85/share. It is now
> down to just above $1. Anyone want to guess how many shares I own?
> 
> Wavelan is a registered trademark of NCR Corporation (I think), formerly
> National Cash Register Corp., which was swallowed by AT&T (I think it was
> before Lucent and ended up in the Lucent spinoff). This was their line of
> very early wireless cards at 1 Mb/s using I think a proprietary protocol.
> 
> Orinoco is a registered trademark of Agere Corp, formerly held by Lucent,
> referring to their line of 802.11b wireless products.
> 
> IEEE is the Institute of Electrical and Electronic Engineers, a USA
> professional society which has an extensive standards activity.  In
> particular, IEEE 802.* governs Ethernet-style packet communications, with
> variants for a broad range of physical media.

Nice history of the companies involved.

As far as the cards themselves go - originally there was the Lucent
WaveLAN IEEE cards.  These were later rebranded as ORiNOCO cards.  At
some point the part of Lucent responsible for the firmware was spun
off into Agere, but (afaik) they don't make cards themselves.
Currently ORiNOCO branded cards and Avaya cards both exist, and are
essentially identical.

The IEEE is responsible for the 802.11 standards.

> > (4) As a matter of (minor) interest,
> > why does the PCMCIA card give a bit-rate of 2Mb/s,
> > while the PCI card claims 11Mb/s ?
> 
> The instances you showed are in Managed mode.  I can think of a few
> reasons:
> 
> a.  The cards are sort of doing managed ad-hoc mode, and aren't getting the
> right packets back from the nonexistent access point, and are responding
> differently.  Cure: set up Ad-Hoc mode properly.
> 
> b.  You're associated with the neighbor's AP, and the PCI card has a good
> signal path while the laptop is behind an obstruction.
> 
> c.  If you see this in Ad-Hoc mode, there's a notorious problem with
> Intersil firmware (Linksys, D-Link, NetGear, etc.)  Using the orinoco
> driver suite, it used to work fine, but now the cards fail to send 802.11b
> ACK packets, so their partners think the link has gone bad, and they reduce
> the speed to compensate, besides retrying transmissions a zillion times.
> In the worst case, throughput goes down to 21 Kbyte/sec, hiss, boo!  It's
> possible that there's no blame in the orinoco driver, but a firmware
> upgrade to 1.0.3 (Intersil) is the cause.

In addition to these, we have had various problems with the reporting
and setting of the bitrate.  This again is a lack of documentation
problem.

> > (5) I always get the warning about different versions
> > (12 against 14 in this case).
> > Does it really matter?
> > And if so, what can one do about it.
> 
> Definitely start reading at:
> http://www.hpl.hp.com/personal/Jean_Tourrilhes/Linux/
> 
> He discusses a lot about wireless LANs. From the subdirectory on wireless
> tools, download the latest version, compile it, and install it. It's also
> possible that your Linux distro has a precompiled updated version in RPM
> format, which would be more palatable when you go to upgrade your distro;
> dig around on the FTP site for the most recent distro version.
> 
> The statistics block given by a driver compiled with wireless tools version
> 14 is not exactly backward compatible, so a version 12 iwconfig will exit
> with a segmentation fault.  It doesn't seem to do any harm, but still, it's
> a good idea to keep the driver and tools at the same version level.
> 
> Good luck!
> 
> James F. Carter          Voice 310 825 2897    FAX 310 206 6673
> UCLA-Mathnet;  6115 MSA; 405 Hilgard Ave.; Los Angeles, CA, USA  90095-1555
> Email: jimc at math.ucla.edu    http://www.math.ucla.edu/~jimc (q.v. for PGP key)
> 

-- 
David Gibson			| For every complex problem there is a
david at gibson.dropbear.id.au	| solution which is simple, neat and
				| wrong.
http://www.ozlabs.org/people/dgibson