The MPLS WG Archive

[Jean Philippe Vasseur <jvasseur@cisco.com>]

Hi Philip,

At 11:45 01/04/2003 -0500, Philip Matthews wrote:

> Jean Philippe: 
>
> Here are some comments on your document:
> draft-vasseur-mpls-nodeid-subobject-00.txt. 
>
> 1) It wasn't clear to me whether this draft is proposing
> that a router push 
>    both its interface address and its nodeid, or
> just push its nodeid. 
>
>    I personally feel that both addresses should be
> included, 
>    as the interface address information can be
> useful for debugging. 

That's perfectly correct.

>    In particular,
> interface information is useful when there are parallel 
>    links between two routers and the links are not
> bundled for some reason. 
>    (The operator might do this, for example,
> because he/she wants to be able 
>    to explicitly specify which link to take when
> routing LSPs between the 
>    two routers. Or, because the links are in two
> different SRLGs and the operator 
>    wants to expose that fact. Or, because the
> links run between routers from 
>    different vendors, and POS bundling has not yet
> been standardized). 
>
> 2) On the assumption that both addresses should be included,
> then I am not sure 
>    that a bit in the Flags field should be used to
> distinguish the two types 
>    of objects. My concern is that an
> implementation that does NOT understand 
>    your new flag might get confused when parsing
> the RRO. The implementation 
>    probably would not break, but if it tries to
> count the number of routers in 
>    the path, or give the operator a more
> "parsed" view of the path, then it 
>    would get the wrong impression.

Well I don't think so, by the way counting the number of RRO IPv4 subobject does not provide any information of the number of hops ...  
If the RRO is used for debugging purposes, this flag will help in giving a more accurate info.

>    My suggestion is to use a new value in the Type field instead. That way, 
>    routers that did not understand this subobject would just skip over it. 
>
> Hope these comments are clear (and useful) ... 

Sure, I just think that we should keep the flag instead of specifying new type.

Thanks for your comment.

JP.

> - Philip 
>