On 1/13/13 12:12 AM, Mikael Abrahamsson wrote:
On Sat, 12 Jan 2013, Matthew Petach wrote:
Thank goodness ethernet never has problems with negotiation going
awry, and coming up with mismatched duplexes, and vendors never had
to implement "no negotiation-auto" in their configs because you
couldn't count on everyone's implementations working together just
absolutely perfectly the first time on bootup. Yes, it sure is a
good thing ethernet never has issues like that which would cripple
your ability to get a box up and running at 2am.
Deployments I've worked on are order of 10^6 managed ports at a time
10^5ish oob ports, auto-negotiation on copper is not a problem that
figures in rollouts anymore and hasn't for more than half a decade.
Has this happened to you with equipment designed and manufactured the
past 5 years?
For me this was a problem with equipment released around 2000, since
2005-2007 or so I haven't seen a single problem that I can recall.
I blame part of this problem on Cisco who was especially bad at
handling autoneg. I remember in 1998 when we couldn't even get link up
between a 100 meg LE interface on a Sun and a (I believe) 3500XL. We
had to use a hub in between to get link at all. Even worse, different
port blocks on the 3500XL behaved differently.