The MPLS WG Archive

[Bob Thomas <rhthomas@cisco.com>]

> In the LDP document "draft-ietf-mpls-ldp-11.txt", there are two
> different failure detection mechanisms.  One is the requirement for
> periodic Hello messages, and the other is the requirement for periodic
> KeepAlive messages in the TCP session.
> 
> What is the reason for requiring both of these mechanisms?  Either one
> alone should be sufficient to detect failures.
> 
> For that matter, why not just use the TCP keepalive option and let
> connections remain up until the TCP connection breaks?
> 
> If this has already been discussed in the past, a pointer to a document
> or web page containing an explanation will be sufficient.

This has been discussed in the past.  You could try wading through:

  http://cell.onecall.net/cell-relay/archives/mpls/mpls.index.html

Bob