Re: [tsvwg] residential broadband BCP PHB and CP treatment Re: CC/bleaching thoughts for draft-ietf-tsvwg-le-phb-04

[Mikael Abrahamsson <[e-mail]>]

On Mon, 23 Apr 2018, [e-mail] wrote:

> [RG] RFC8100 should explain that. If RFC8100 requires clarifications, 
> please point out the relevant issues.

RFC8100 has TL;DR; problems for my use-case. It also doesn't give any 
concrete operational advice.

If someone comes to me and asks "hm, what service-policy should I 
configure towards my 500 internet peers and transit, and towards my 
residential access customers?" and I point them to RFC8100, do you think 
they'd be much wiser even if they forced themselves to read it all?

If they come to me and ask "oh, I want to IPv6-enable my home gateways, 
what should I think of" I can point them to RFC7084 which doesn't have any 
"TL;DR;" issues at all, it's all relevant to their question.

I want to create a BCP (or informational) document that gives concrete 
advice on what to do when you want to support Internet-facing QoS that 
might be slightly better than single FIFO and "set CP=0" (or ignoring it). 
RFC8100 will probably be referenced by this document, but most likely very 
few will read RFC8100 in the process of implementing what they want to 
implement.

Today the policy towards peers/IXes/transits probably looks like this if 
you're running a cisco-router (hand-typed so might not apply cleanly):

policy-map ingress-from-internet-peers
   class class-default
     set dscp 0

Potentially if they run a MPLS-enabled network, they might use "set mpls 
experimental imposition 0" and then just ignore the DSCP bits in their 
MPLS network. Similar towards residential customers.

I want to offer an alternative that's still sane trade-off between risks 
of DDOS-attacks from the Internet, and also that supports LE.

-- 
Mikael Abrahamsson    email: [e-mail]