Re: revision of architecture draft is now published

[marcelo bagnulo braun <marcelo@it.uc3m.es>]

> However I suppose that the question here is that given that some 
> considerations about TE should be included in the document, what text 
> is appropriate here that notes the goal, but references the large body 
> of work that has been undertaken already on QoS and TE?
>
> Suggestions are, obviously, solicited!

Well, after re-reading the document, it is my impression that the 
document actually already contains several references to this point.
First, the document includes it explicitely in the base scenario

Section 2:

  In addition there is the potential consideration of being able to
  distribute traffic load across a number of network paths according to
  some pre-determined objective, as a form of traffic engineering.

which is all that imho is needed in this part of the doc.

After, in section 4.2 the document describes the key point of how path 
selection is affected in this new multiaddress configuration

Section 4.2

  The further implication here is that path selection (ISP A vs ISP B
  transit for incoming packets) is an outcome of the process of
  selecting an address for the destination host.

Moreover, in the discussion of section 5.3.1 several considerations 
about locator selection are presented.
Perhaps an option would be to include a new subsection in section 5.3 
about policy based locator selection.
In this section, it could be possible to highlight that in this new 
multiaddress configuration, policy (and TE in general) is no longer 
solely achieved through manipulation of the routing system, but a 
relevant part of the path selection is performed through address 
selection and that it is possible to achieve policing and TE by 
influencing locator selection process. In particular a reference RFC 
3484 and its policy table could be useful.

Another interesting point that could be considered in the analysis is 
that the element that performs the locator selection (host or edge 
router) determines also policy, so the choice of using packet rewriting 
or host based locator selection shifts the policy capability from one 
element to the other.
this is relevant imho, because if hosts perform policing, then a more 
fine grained policy is simpler to achieve, because hosts can select 
locators depending for instance in the application that is sending 
traffic. However, policy configuration in every host may be a 
management problem. Finally, policy enforcement may be rather 
challenging when the host based approach is used, since it is up to the 
hosts themselves. In the edge router case, policy enforcement by router 
is more similar to what we have today imho.

well, this kind of stuff is what i think could be interesting to 
include in the analysis, since may be relevant for instance when 
selecting between the host based and the edge router based locator 
selection.

Regards, marcelo






> thanks,
>
>   Geoff