Re: [tsvwg] CC/bleaching thoughts for draft-ietf-tsvwg-le-phb-04

[Mikael Abrahamsson <[e-mail]>]

On Thu, 19 Apr 2018, Bless, Roland (TM) wrote:

> Ok, so _bleaching_ for me would be to set the DSCP to 0. The subsequent

Ok, what should I call the following (all values decimal for the 6 DSCP CP 
bits)

Look at diffserv CP values 0-63.
If CP in range 2-63, SET CP=0
If CP == 1 pass unchanged
If CP == 0 pass unchanged

This bleaches 2-63 to 0 and leaves 1 unchanged. What do we call this?

> This is a bit confusing. What do you mean by "bleach" here, treat
> LE as BE? Remarking of DSCPs is a somewhat orthogonal problem since

For me there are two things an ISP does depending on DSCP values. It has a 
policy that (might) inspect CP values and change them. It also has a PHB 
depending on the CP values.

The simplest policy is "SET CP = 0 on all packets". This seems to be what 
you call bleaching.

> Diffserv requires a flexible mapping from DSCP -> PHB. So my
> recommendation is to remark CS1 to LE for domains that formerly used
> CS1 as LE codepoint. Otherwise, you cannot be sure whether CS1 actually
> is CS1 in its original meaning or whether it should be treated as LE.
> Removing that ambiguity is the main motivation for my LE PHB draft.

If the domain previously just bleached everything and now wants to deploy 
support for LE (both allowing the CP to enter the ISP network and also 
creating different PHB for LE vs BE), what should it do regarding CS1?

Saying "it depends" is not helpful. This is the Internet with billions of 
devices and users. What should be the default behaviour for non-managed 
devices run by normal people that they bought from the electronics store. 
I am not talking about managed enterprise networks etc. I am talking about 
the most common network out there, the "oh, I need Internet access to 
watch cat pictures on my own device and it's default out of box configured 
from DSCP point of view".

>> 3. Bleach rest to BE.
>
> What is the rest?

CP between 2 and 63. Not 0 and 1.

-- 
Mikael Abrahamsson    email: [e-mail]