The MPLS WG Archive

[Vach Kompella <vkompella@JasmineNetworks.com>]

Eric,

 

OK, I didn't get: you were basically restating my point.  However, I think you don't resolve such ambiguities in BCPs.  Changes in the text of the draft (or more explicit EROs) are more apropos.

 

-Vach

-----Original Message-----
From: Eric Gray [mailto:eric.gray@sandburst.com]
Sent: Tuesday, February 27, 2001 1:22 PM
To: Vach Kompella
Cc: Kireeti Kompella; yakov@juniper.net; mpls@UU.NET
Subject: Re: Expanded ERO draft

Vach,

    Thanks for the prompt reply.  However, it doesn't really
help.  :-)

    Consider that I may have a function "isMe()" that I use
to eliminate self-references in an ERO.  If, as you say, B
recognizes that '1' is an interface by which upstream LSR
A knows B, then this function should return true every time
it is passed the value '1' in a context in which it has gotten
an ERO in a message from LSR A.  In this case, iterative
calls to "isMe()" would result in removing all of the '1's
in the ERO.

--
Eric Gray

You wrote:

 Eric,The interfaces shown are numbered with an outbound perspective, if I understand Kireeti's picture correctly.  B understands that the AB link is numbered 1 according to A, and that its own BC link is numbered 1.My comment is on clarifying the processing of the unnum subobjects.-Vach

-----Original Message-----
From: Eric Gray [mailto:eric.gray@sandburst.com]
Sent: Tuesday, February 27, 2001 12:48 PM
To: Kireeti Kompella
Cc: vkompella@JasmineNetworks.com; yakov@juniper.net; mpls@UU.NET
Subject: Re: Expanded ERO draft
 

Kireeti,

    Out of curiosity, how many un-numbered interfaces can
B have that are known to it as interface 1?

--
Eric Gray

You wrote:

Vach, Yakov,

> > The other aspect is the first clause of the quoted sentence above from the
> > rsvp-unnum draft: "all initial subobjects that refer to itself."
>
> The text needs to be clarified to say something like "all initial subobjects
> of type other than Unnumbered Interface ID"...".

While the change in text is possible, it is clear that, unless the
interface is physically looped back to itself, the second subobject
does *not* refer to "itself".

             1       1       1
         A ----- B ----- C ----- D

B receives an ERO of (1, 1, 1) from A (PHOP).  B checks that (A, 1)
refers to itself, removes it.  Now it sees unnum interface 1, which
clearly points at C.  Why would B think that this "1" refers to B?

Kireeti.