Re: [v6ops] draft-pauly-v6ops-happy-eyeballs-update-01

[Philip Homburg <[e-mail]>]

>On Thu, 30 Mar 2017, Philip Homburg wrote:
>
>> It is hard to quantify, but in IPv4 lots of potential PMTU problems are 
>> masked by MSS clamping.
>
>Yes. I don't have data, but I have commonly talked to people in the 
>business who are of the opinion that MSS clamping solves "all" problems.
>
>When I point out that there are other protocols I typically get back that 
>MSS clamping is good enough and solved all their customer complaints.

It's sad how little some people know about internet protocols. 

>While I don't really like the approach, I think the pragmatic thing to do 
>is to have the devices themselves have the ability to do PMTUD clamping if 
>they have lots of problems that can be attributed to PMTUD, and then have 
>this reported by means of diagnostics report to application/device vendor.

The 'nice' thing about doing MSS clamping in a CPE is that the CPE knows 
what it is using to connect to the internet. 

A host would have to probe every time it connects to a network for local
PMTU limitations. I wonder what the operational impact would be if every
host would start sending maximum sized packets to some well known destination
right after connecting to the network.