IP Header Compression

Voir toute la rfc dans une seule page

Télécharger le PDF

RFC 2507                 IP Header Compression             February 1999


   This method is essentially the differential encoding techniques of
   Jacobson, described in [RFC-1144], the differences being the placement
   of the compressed TCP header fields (see section 6), the use of the
   O-flag, the use of the R-flag, and elimination of the C-flag.  The
   O-flag allows compression of the TCP header when the Timestamp option
   is used and the Options fields changes with each header.

   DELTA values (except for Reserved field and Options, Padding) MUST be
   coded as in [RFC-1144].  A Reserved field value passed with the R-flag
   MUST NOT update the context at compressor or decompressor.

7.12.2. Without differential encoding

       Source Port           NOCHANGE  (DEF)
       Destination Port      NOCHANGE  (DEF)

       (all the rest)        RANDOM

   The Identification field in a preceding IPv4 header is RANDOM.

   A packet with a TCP header compressed according to the above must be
   indicated to be of type COMPRESSED_TCP_NODELTA.  It uses the same CID
   space as COMPRESSED_TCP packets, and the header MUST be saved as
   context. The compressed header is described in section 6.

   This packet type can be sent as the response to a header request
   instead of sending a full header, can be used over links that reorder
   packets, and can be sent instead of a full header when there are
   changes that cannot be represented by a compressed header. A
   sophisticated compressor can switch to sending only
   COMPRESSED_TCP_NODELTA headers when the packet loss frequency is high.




















Degermark, et. al.          Standards Track                    [Page 32]