RIPngプロトコル適用性報告

   This memo provides information for the Internet community.  This memo
   does not specify an Internet standard of any kind.  Distribution of
   this memo is unlimited.

   As required by Routing Protocol Criteria (RFC 1264), this report
   defines the applicability of the RIPng protocol within the Internet.
   This report is a prerequisite to advancing RIPng on the standards
   track.

   The RIPng protocol description is defined in RFC 2080.

   This report describes how RIPng may be useful within the new IPv6
   Internet.  In essence, the environments in which RIPng is the IGP of
   choice is comparable to the environments in which RIP-2 (RFC 1723) is
   used in the IPv4 Internet.  It is important to remember that RIPng is
   a simple extrapolation of RIP-2; RIPng has nothing conceptually new.
   Thus, the operational aspects of distance-vector routing protocols,
   and RIP-2 in particular, within an autonomous system are well
   understood.

   It should be noted that RIPng is not intended to be a substitute for
   OSPFng in large autonomous systems; the restrictions on AS diameter
   and complexity which applied to RIP-2 also apply to RIPng.  Rather,
   RIPng allows the smaller, simpler, distance-vector protocol to be
   used in environments which require authentication or the use of
   variable length subnet masks, but are not of a size or complexity
   which require the use of the larger, more complex, link-state
   protocol.

   The remainder of this report describes how each of the features of
   RIPng is useful within IPv6.

   A goal in developing RIPng was to make the minimum necessary change
   to RIP-2 to produce RIPng.  In essence, the IPv4 address was expanded
   into an IPv6 address, the IPv4 subnet mask was replaced with an IPv6
   prefix length, the next-hop field was eliminated but the
   functionality has been preserved, and authentication was removed.
   The route tag field has been preserved.  The maximum diameter of the
   network (the maximum metric value) is 15; 16 still means infinity
   (unreachable).

   The basic RIP header is unchanged.  However, the size of a routing
   packet is no longer arbitrarily limited.  Because routing updates are
   never forwarded, the routing packet size is now determined by the
   physical media and the sizes of the headers which precede the routing
   data (i.e., media MTU minus the combined header lengths).  The number
   routes which may be included in a routing update is the routing data
   length divided by the size of a routing entry.

   The address field of a routing entry is 128 bits in length, expanded
   from the 32 bits available in RIP-2.  This allows the RIP entry to
   carry an IPv6 prefix.

   The 32-bit RIP-2 subnet mask field is replaced by an 8-bit prefix
   length field.  It allows the specification of the number of bits in
   the prefix which form the actual prefix.

   The ability to specify the next hop, rather than simply allowing the
   recipient of the update to set the next hop to the sender of the
   update, allows for the elimination of unnecessary hops through
   routers which are running multiple routing protocols.  Consider
   following example topology:

         -----   -----         -----   -----
         |IR1|   |IR2|         |XR1|   |XR2|
         --+--   --+--         --+--   --+--
           |       |             |       |
         --+-------+-------------+-------+--
           |--------RIPng--------|

   The Internal Routers (IR1 and IR2) are only running RIPng.  The
   External Routers (XR1 and XR2) are both running BGP, for example;
   however, only XR1 is running BGP and RIPng.  Since XR2 is not running
   RIPng, the IRs will not know of its existance and will never use it
   as a next hop, even if it is a better next hop than XR1.  Of course,
   XR1 knows this and can indicate, via the Next Hop mechanism, that XR2
   is the better next hop for some routes.

   Authentication, which was added to RIP-2 because RIP-1 did not have
   it, has been dropped from RIPng.  This is safe to do because IPv6,
   which carries the RIPng packets, has build in security which IPv4 did
   not have.

   By allowing RIPng routing update packets to be as big as possible,
   the number of packets which must be sent for a complete update is
   greatly reduced.  This in no way affects the operation of the
   distance-vector protocol; it is merely a performance enhancement.

   The limit of 15 cost-1 hops is a function of the distance-vector
   protocol, which depends on counting to infinity to resolve some
   routing loops.  If infinity is too high, the time it would take to
   resolve, not to mention the number of routing updates which would be
   sent, would be prohibitive.  If the infinity is too small, the
   protocol becomes useless in a reasonably sized network.  The choice
   of 16 for infinity was made in the earliest of RIP implementations
   and experience has shown it to be a good compromise value.

   RIPng will efficiently support networks of moderate complexity.  That
   is, topologies without too many multi-hop loops.  RIPng also
   effeciently supports topologies which change frequently because
   routing table changes are made incrementally and do not require the
   computation which link-state protocols require to rebuild their maps.

   Because the basic protocol is unchanged, RIPng is as correct a
   routing protocol as RIP-2.  RIPng serves the same niche for IPv6 as
   RIP-2 does for IPv4.

   RIPng security is discussed in section 3.4.

   Gary Scott Malkin
   Xylogics/Bay Networks
   53 Third Avenue
   Burlington, MA 01803

   Phone:  (617) 238-6237
   EMail:  gmalkin@xylogics.com