Re: [Openl2tp-users] Tie breaker AVP and peer collision.

[James C. <jch...@ka...>]

Karthik Ravikanti wrote:
> Hi guys,
> 
> 1) Is the tiebreaker AVP included to ensure that there exists one and
> only one tunnel between the peers?
> 
> 2) Or is it included to ensure that not more than one tunnel is
> negotiated at the same time?
> 
> 1) Is how openl2tp seems to be implementing it. 2) Is what RFC2661 seems
> to imply.

It is 1). Here is what rfc2661 says:-

 Tie Breaker (SCCRQ)

      The Tie Breaker AVP, Attribute Type 5, indicates that the sender
      wishes a single tunnel to exist between the given LAC-LNS pair.

> I even cross-checked other implementations. Everybody seems to have
> their own interpretation about it.
> 
> Looking at L2TPv3 my belief in 2) got reinforced. Please reply with your
> comments.

In L2TPv3, there is a Control Tiebreaker AVP and a Session Tiebreaker
AVP. The Control Tiebreaker is identical to the L2TPv2 version and
behaves the same way. The Session tiebreaker AVP is new in L2TPv3 and is
used to detect and handle two sessions being set up at the same time.

However, the attribute number of L2TPv3's Session Tiebreaker AVP is 5,
which is the same value used for L2TPv2's Tiebreaker AVP and L2TPv3's
Control Tiebreaker AVP. The L2TPv3 protocol treats attribute 5 as a
tunnel or a session tiebreaker depending on whether it is present in a
tunnel message or a session message. Perhaps this is why confusion has
arisen?

> -- 
> Karthik Ravikanti
> Associate Engineer,
> Primesoft I.P. Solutions
> Block III, White House, Begumpet
> Hyderabad.
> http://students.iiit.net/~karthik_ravikanti

-- 
James Chapman
Katalix Systems Ltd
http://www.katalix.com
Catalysts for your Embedded Linux software development