[KLUG Members] Re: Linux compatible NIC card

[Bryan J. Smith]

On Fri, 2002-04-12 at 23:05, Adam Williams wrote:
> Linksys, Dlink, 3Com, Intel.  Just about anything not off-the-wall will
> work.  I personally avoid Linksys as they have made some real lemons in
> the past (cards with a different MAC every time they power up, etc..)

Regarding Linksys, that's not their problem.  It's a problem in the
splintered Digital 2114x-series (tulip) market.  Blame Palmer of Digital
for selling off that division which powered 13 of the top 16 10/100 NICs
at one point (circa 1996).  The "tulip" has splintered into hundreds of
variants as a result.

In Linksys' defense, they *DO* change the version number each time they
change the ASIC, but I think they should change the model number
instead.  NetGear has also been guilty of doing this and, _unlike_
Linksys, they don't print the NIC's "revision number" on the box like
Linksys.

DLink uses the RealTek chips which have a higher CPU utilization rate
and don't always support Full Duplex.  As such, I wouldn't use them in
servers.  The SiS900 is a variant of the RealTek series that requires a
major driver rewrite (although SiS has been supporting development
directly with a GPL driver in newer kernels -- the one in 2.4.18 seems
near-perfect now).

3Com seems to be solid now, since they are supporting their products
directly.  You usually pay for this in the board costs though.  Like
DLink, 3Com also has a nasty history of changing ASICs on the board
without printing the revision number on the box.  Since 3Com's NICs have
_never_ been tulip-based, this is 100% of their own doing.

Intel is a catch-22.  They don't publish a lot of specs, so the GPL
drivers usually have issues _until_ someone reverse engineers some
undocumented "features."  As such, if you use the non-GPL, Intel OEM
drivers (i.e. the "e100" and "e1000"), you're much better off.  Luckily
most distros that highly customize their kernels, e.g. RedHat and SuSe,
include those drivers in their kernels.

-- Bryan

-- 
The USDOJ v. Microsoft trial will result in unconditional surrender.
No matter who wins, the consumer will be subject to the victor's
"terms."  Which is worse?  Clueless government or clueless monopoly?
--------------------------------------------------------------------
Bryan J. Smith, SmithConcepts, Inc.        mailto:b.j.smith@ieee.org
Engineers and IT Professionals          http://www.SmithConcepts.com