There is an I-D with the same name as an RRG proposal - Christian
Vogt's Name Based Sockets (NBS):

  http://tools.ietf.org/html/draft-ubillos-name-based-sockets-02

   Javier Ubillos, Mingwei Xu, Zhongxing Ming and Christian Vogt

I haven't read it all, but my impression is that it involves a new API
by which applications can attempt to communicate with other hosts
using NBS, without first knowing that the other host is capable of
doing so.  This would require extra capabilities in the stack and
significant changes to existing applications.

There's no statement of goals and non-goals, but I think it aims to
provide multihoming and mobility, and in so doing probably would
provide portability of host identities between the PA prefixes of
particular ISPs.

The name information is piggybacked onto otherwise normal packets, at
least for the initial one or few packets, by way of an IPv4 header
option or an IPv6 extension header.  I recall from previous RRG
discussions that packets with IPv4 header options are likely to be
handled by many routers, including DFZ routers, in the "slow path" -
that is by a CPU rather than the main FIB forwarding mechanism.  So I
guess this makes NBS impractical or unattractive in today's IPv4 Internet.

The I-D involves changes to Shim6 which would be part of NBS -
implicitly to provide multihoming - and with some further changes, to
support mobility.


I think that to implement NBS, there would need to be changes to
application code, probably changes to application protocols, and new
functionality in stacks.  I think this makes it impossible to
contemplate as an overall solution to the routing scaling problem,
unless perhaps it was assumed that the IPv4 routing scaling problem
was not important, and if somehow most or all applications on IPv6
would use NBS.  The I-D contains no mention of the routing scaling
problem, or claims that it would be a general solution to it.

  - Robin

_______________________________________________
rrg mailing list
[email protected]
http://www.irtf.org/mailman/listinfo/rrg

Reply via email to