Re: [tsvwg] LE PHB -01 draft comments (from David Black)

[<[e-mail]>]

Hallo Roland,

the draft reads better now. David commented version 01 and I'd like to pick up some of his content related comments (marked [DB]) and add some own ones (marked [RG]).

Regards, Ruediger

-- Section 1

[DB] The last bullet in Section 1.1. contains some discussion of "downgraded" 
that goes beyond an example, and ought to be merged into the last paragraph 
on p.3.

[RG] I share David's point of view.

-- Section 1.2

[RG]
OLD:
      No harm to other traffic: since the LE PHB has got the lowest
      forwarding priority it does not consume resources from other PHBs.
      Deployment across different provider domains causes no trust
      issues or attack vectors to existing traffic.

NEW
      No harm to traffic of other PHBs: since packets marked for LE PHB 
      are dropped prior to packets of any other PHB, they don't consume 
      resources required by other PHBs. Deployment across different 
      provider domains causes no trust issues or attack vectors to 
      traffic of other PHBs.
-----------------------
[DB] I wonder whether the 2nd and 3rd bullets are overstated.  There is some 
network switch/router configuration required and something like a low priority 
queue has to be set up to drop LE traffic when there are other competing uses.   
I don't think the bullets are actually wrong, but some additional text to point 
out what does have to be done to use the LE PHB would be helpful.

[RG] I share David's point of view.


-- Section 2
[DB]
OLD
   Ideally, LE packets SHOULD be forwarded only if no best-
   effort packet is waiting for its transmission.
NEW
   Ideally, LE packets SHOULD be forwarded only if no packet
   with any other PHB is awaiting transmission.

[DB] Understated - should apply to all PHBs, not just default.

[RG] Agree with David.
----------
[RG] There are more options to produce LE: a router may offer 
a committed/excess traffic configuration. committed rate/weight 
could be 0 for LE, only an excess rate/weight may be configured. 
If the default PHB has a committed rate above 0 and the default 
PHB excess weight is bigger than LE, LE shouldn't be able to 
seize capacity from default PHB.
If RED is deployed, Default PHB and LE may be produced using 
the same queue. All LE traffic should be dropped by RED prior 
to the first Default PHB packet.


--Section 5

[RG] I wonder, whether ECN related aspects should be discussed:
- should any LE traffic be dropped prior to ECN marks for any 
  other PHB? I think, that is true if ECN is used to indicate 
  congestion.
- should LE be able to support ECN? Or are LE packet drops the 
  only reaction in the case of congestion.

-- Section 8

[RG] Unauthorised marking of traffic as LE may be regarded as 
a kind of denial of service. 
On the other hand, may re-marking traffic which is assumed to 
be dDoS traffic as LE be an alternative to dropping this 
traffic (I can't judge that, may be in some special cases?).