eepro100 confusion

Donald Becker becker at scyld.com
Mon Nov 11 16:00:17 PST 2002


On Mon, 11 Nov 2002, Florian Gabriel wrote:

> >Cisco switches had a broken autonegotiation implementation.  Rather than
> >fix the deployed hardware, Cisco taught that autonegotiation was
> >unreliable and should be turned off.  OK.  But they went further and
> >recommended forcing full duplex.
> >
> A new document ( http://www.cisco.com/warp/public/473/46.html ) can make
> some light here.

Very interesting!
This is a document dated Nov. 1 2002 -- apparently just released.

The vital quote: 

    There are many opinions on the subject of auto-negotiation.
                   ^^^^^^^^^
    Previously, many engineers were advising customers not to use
    auto-negotiation with any switch connected device. However,
    improvements in the interoperation of auto-negotiation and the
    maturity of the technology has recently changed the view of using
    auto-negotiation. In addition, performance issues due to duplex
    mismatches as a result of manually setting speed and duplex only on
    one link partner, have become more common. Due to these recent
    issues, using auto-negotiation is regarded as a valid practice.
                                                    ^^^^^^^^^^^^^^

Uhmm, "it works" isn't really an "opinion"!  I guess this one way of
admitting they were wrong.  But by waiting so long, they now they now
have a major education problem.

"Remember that thing we said was evil before?  Uhmmm, now it's good.
Oh, and our new firmware defaults to turning it on and breaks all of
your existing installations.  Why did we do it?  Because Donald really
needed twenty or thirty extra emails a day."

-- 
Donald Becker				becker at scyld.com
Scyld Computing Corporation		http://www.scyld.com
410 Severn Ave. Suite 210		Scyld Beowulf cluster system
Annapolis MD 21403			410-990-9993




More information about the Beowulf mailing list