The MPLS WG Archive

Cell Relay Retreat>MPLS WG Archive>month:2000-Nov> msg00048



[Date Prev][Date Next][Thread Prev][Thread Next]  
  [Date Index][Thread Index][Author Index][Subject Index]

Fatal Errors in LDP

  • From: Bob Thomas <rhthomas@cisco.com>
  • Date: Tue, 07 Nov 2000 07:28:18 -0500
  • cc: mpls@UU.NET

Abhijit,

> In the LDP draft, following error codes are specified as Fatal Errors. 
> (refer section 3.5.1.2)
> 
> Bad LDP Identifier
> Bad Protocol Version
> Bad PDU Length
> Bad Message Length
> Bad TLV Length
> Malformed TLV value
> 
> An LSR receiving Notification message with above status codes is supposed
> to cleanup the LDP session along with the learned FEC-LabelMappings.. as
> explained in section 3.5.1
> 
> This could lead to session cleanup even if one message is malformed. Is it
> not a bit harsh? Instead can it be so that a count of error notifications
> for sessions be maintained and when the Fatal notifications reach that
> count cleanup the session with the peer. Can an implementation choose to
> do so? 

The rationale for making these particular errors fatal is that when
one of them occurs there is something very seriously wrong and there
is no obvious way the LSR detecting the condition can recover.

Bob