[c-nsp] necessity of nowadays

[Saku Ytti]

On 23 March 2016 at 15:16, Phil Mayers <p.mayers at imperial.ac.uk> wrote:
> I disagree. We ran UDLD on 6500 for 5+ years, and it caused way more
> problems (particularly due to false-positives) than it solved.
>
> We ripped it out, and I've no cause to question that decision.

Particularly annoying is that you can't MLS ratelimit BPDU if you run
UDLD, so you expose yourself to plethora of control-plane congestion
vectors. But if you already MUST run STP, then this option is gone
anyhow, then it's not that bad. But if BPDU is only reason to accept
BPDU, then it certainly is going to cause more pain than solve.

-- 
  ++ytti