The MPLS WG Archive

Cell Relay Retreat>MPLS WG Archive>month:2004-Jan> msg00059



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

draft-ietf-mpls-in-ip-or-gre-04.txt

  • From: Eric Rosen <erosen@cisco.com>
  • Date: Thu, 22 Jan 2004 09:53:16 -0500
  • Cc: zinin@psg.com
  • User-Agent: EMH/1.14.1 SEMI/1.14.3 (Ushinoya) FLIM/1.14.3(Unebigoryōmae) APEL/10.3 Emacs/21.3(sparc-sun-solaris2.8) MULE/5.0 (SAKAKI)

The IESG  has reviewed draft-ietf-mpls-in-ip-or-gre-03.txt, and  have made a
number of  comments.  In  response, I have  issued draft-ietf-mpls-in-ip-or-
gre-04.txt.   I will  briefly review  the  comments below  and indicate  the
changes I've  made in response.  (The raw  comments can be found  in the i-d
tracker.) I  encourage all interested  parties to review the  document again
before it is resubmitted to the IESG.

1. IPv6 compatiblity. 

   It was pointed  out that the draft was not very  specific about IPv6, and
   was also  very sloppy in its  use (or lack thereof)  of IPv6 terminology.
   The new version incorporates proper IPv6 terminology and procedures.

2. Differentiated Services

   The old version said only that the  setting of the MPLS EXP bits could be
   considered  when  determining  how  to   set  the  DS  field  in  the  IP
   encapsulation header, and vice versa.  It was pointed out that this might
   be too  much freedom, as there is  existing work on MPLS  Diffserv and on
   Diffserv  in  Tunnels that  needs  to  be  referenced.  The  new  version
   clarifies  the intention  by  using the  more  rigorous terminology  from
   diffserv and  referring to the necessary diffserv  specifications for the
   details.

3. What one tunnel endpoint must know about the other

   Some text  was added  indicating that  the tunnel head  must know  the IP
   address and tunnel  capabilities of the tunnel tail,  that this knowledge
   could be  obtained by a variety of  means which are outside  the scope of
   this document, and that this  means might depend on the application which
   is using  the tunnel.  This  was added in  response to an  objection that
   we needed to say a bit more than just "tunnel setup is out of scope".

4. GRE optional fields. 

   The draft had an absolute prohibition against the use of the GRE optional
   fields, but did  not give any rationale for this.   An objection was made
   that this absolute prohibition is  not really appropriate.  The draft now
   states that,  by default,  these optional fields  are not used,  but that
   they may be used if both endpoints agree.  There is also some text saying
   why these optional fields are not generally useful.

5. Fragmentation and reassembly. 

   The  draft had  an absolute  prohibition  against the  tunnel head  doing
   fragmentation.  It was objected that this is too strong, that there might
   be  situations  in  which  fragmentation  and reassembly  at  the  tunnel
   endpoints  would be  advisable.  The  draft now  states that  by default,
   there is not to be any  fragmentation, but that an implementation must be
   configurable to allow fragmentation.

6. Security

   The  draft said  that if  the tunnels  could, if  desired, be  secured by
   IPsec.  This raised the usual objection that "it is not sufficient to say
   just use IPsec".  The draft now goes  into a bit more detail about how to
   use it to achieve various security goals.