which BGP4+ used

[kunihiro@zebra.org]

>I know a version of BGP4+ is running over 6bone, but it seems
>different from draft-ietf-idr-bgp4-multiprotocol-01.txt.  I'm now
>testing our BGP4+ code with cisco at one of 6bone backbone sites,
>but its BGP4+ is apparently different from the draft. I'm afraid
>I might miss something new. Is there any updated spec?

I'm thinking about a same problem.  It seems that some current running
code conforms draftr-ietf-idr-bgp4-multiprotocol-00.txt (include two
octet NLRI length field).  Many people are thinking that it is also
important conforming current vendor's beta version implemetation.  But
I'm not sure vedor implementation.

I have seen three types of BGP4+ packet.  Below description assumes
SNPA is zero.

1. o gated-bgp4+ default (without -DBATES_LAST_DRAFT)
  .. no SAFI 
  +--------------------------------+
  | AFI (2 octets)                 |
  +--------------------------------+
  | Length of NextHop (1 octet)    |
  +--------------------------------+
  | Address of NextHop (variable)  |
  +--------------------------------+
  | Number of SNPAs (1 octet = 0)  |
  +--------------------------------+
  | NLRI Length (2 octets)         |
  +--------------------------------+
  | NLRI (variable)                |
  +--------------------------------+

2. draft-ietf-idr-bgp4-multiprotocol-00
   o gated-bgp4+ (with -DBATES_LAST_DRAFT )
   o MRT
   o zebra (with draft-00 option)
   .. with NLRI length
  +--------------------------------+
  | AFI (2 octets)                 |
  +--------------------------------+
  | SAFI (1 octet)                 |
  +--------------------------------+
  | Length of NextHop (1 octet)    |
  +--------------------------------+
  | Address of NextHop (variable)  |
  +--------------------------------+
  | Number of SNPAs (1 octet = 0)  |
  +--------------------------------+
  | NLRI Length (2 octets)         |
  +--------------------------------+
  | NLRI (variable)                |
  +--------------------------------+
  
3. draft-ietf-idr-bgp4-multiprotocol-01
   o zebra
  ... rid of NLRI length
  +--------------------------------+
  | AFI (2 octets)                 |
  +--------------------------------+
  | SAFI (1 octet)                 |
  +--------------------------------+
  | Length of NextHop (1 octet)    |
  +--------------------------------+
  | Address of NextHop (variable)  |
  +--------------------------------+
  | Number of SNPAs (1 octet = 0)  |
  +--------------------------------+
  | NLRI (variable)                |
  +--------------------------------+

IMHO, the best thing is moving toward to draft-ietf-idr-bgp4-multiprotocol-01
specification.
--
Kunihiro Ishiguro