[c-nsp] Confirmation of Gigabit Ethernet autonegotiation behavior

[Saku Ytti]

On (2013-01-25 12:01 +0000), Wayne Lee wrote:

> > I don't think it's technical TBH. I suspect it's just "telco" mindset -
> > "force all the params to on/fast/full and it's better, right?"
> 
> Virgin do the same thing.

My employer does the same for most products and often in internal links.
It's certainly due to legacy reasons which just have been inherited, I've
not found anyone who passionately wants them to be like that but changing
stuff is hard.

Always demand from your provider auto-nego in your RFQ/RFP, this will help
lot, often internally techies asking for stuff is hard to sell, business
driven demands are much easier to met.

Good selling point for autonego is RFI or remote-fault-indicator which will
help put both sides of the link down, which will help with some blackholing
issues you'll have when running without autonego.

Other selling point would be 'operative down', which autonego standard
allows to signal and chips support, but is not capitalized by switch or
router operating systems. So you could differentiate in syslog messages
when remote end went down for 
  a) because far end punched 'shutdown'
  b) some other reason

When you see a) you'll save lot of time and money in debugging.

-- 
  ++ytti