[c-nsp] 3508 / GBIC ZX / Minimum IOS

[Richard A Steenbergen]

On Tue, Jan 09, 2007 at 06:00:21PM -0500, Deepak Jain wrote:
> All I can say (and I may be bitter) is that if autoneg on GigE was so 
> grand... um... why has nearly every SP found reason to disable it?? And 
> shouldn't auto neg (without neg on the other side) default to the same 
> state on both sides?? -- you know, put the "auto" in "autoneg"??

In all honesty, autonegotiation problems have been effectively gone for a 
very very long time now. The real problems were in the first generation 
implementations on FastEthernet, and then a handful of problems with GigE 
when it first came out. The continued desire to disable auto-negotiation 
is senseless and a knee-jerk reaction, but SP's still do it because its 
been engrained into them. In a production state auto-negotiation is 
actually left enabled, since it is capable of detecting many "one side of 
a bidirectional link failure" issues. In a diagnostic state however, for 
gear without light meters onboard you'll get link if you are neg disabled 
regardless of the status of negotiation on the remote side, so many people 
still use it for that.

Fortunately the 10GE spec did what the 1GE spec should have done, and 
cleaned house of legacy things which no longer have a place in modern 
networks like "no negotiation" and "half duplex". :)

-- 
Richard A Steenbergen <ras at e-gerbil.net>       http://www.e-gerbil.net/ras
GPG Key ID: 0xF8B12CBC (7535 7F59 8204 ED1F CC1C 53AF 4C41 5ECA F8B1 2CBC)