[c-nsp] does duplex mismatch affect UDP throughput?

[Mikael Abrahamsson]

On Tue, 2 Aug 2011, Reuben Farrelly wrote:

> Not to mention it also breaks MDI-X... grrr.

It doesn't really, just on some platforms. Just the same way that there is 
absolutely no reason for the device to stop advertising autoneg 
capabilities just because 100/full was forced, is there a reason to turn 
off MDI-X just because duplex and speed was forced. This is a matter of 
implementation. I feature requested continuing of autoneg being on even 
though speed and duplex was forced, to Cisco and other vendors 6-8 years 
ago, and got very little traction back. I'm told some catalyst switches 
nowadays have this feature.

I encourage everybody to require autoneg and auto MDI/MDX to stay on when 
you do RFQs. It's time this operational nightmare was put out of its 
misery.

-- 
Mikael Abrahamsson    email: swmike at swm.pp.se