Re: shim6 @ NANOG (forwarded note from John Payne)

["Per Heldal" <heldal@eml.cc>]

On Mon, 27 Feb 2006 16:13:43 -0500 (EST), "Jason Schiller
(schiller@uu.net)" <jason.schiller@mci.com> said:
[snip]
> I.m suggesting there must be some other way to provide current IPv4 style
> TE capabilities for IPv6 without de-aggregating to the global Internet
> routing table.  This can be in shim6 or 8+8 / GSE, or something else. 
> But
> supporting the full suite of TE capabilities is a day 1 requirement for
> business customers, not just loading all links and failover.

That would mean no v6 for another 5 years or more. It isn't possible to
implement something that doesn't exist. How do you sell that
idea to the .gov crowd and others who are making serious plans for
migration within the next 2-3 years?

An alternative could be "controlled fragmentation" through temporary PI
assignments to be re-evaluated periodically and to be revoked once the
technology is ready. Renumbering of end-sites should be a non-issue once
all the supporting software and hardware gadgets are in place. [This is
probably an issue for the RIR-communities + IANA to work out.] 

Personally I see this primarily as an alternative if v6 routing isn't
solved before we run out of v4 addresses. It doesn't make that much
sense to implement v6 as a sort of "v4+".

> 
> Is it the consensus of the shim6 working group that the full suite of TE
> capabilities should not be a requirement?  Or is this just the opinion of
> a few vocal people?
> 

Don't researchers picture a slightly more modular approach? The basics
of shim6 must have "hooks" for TE, but it is a long way from there to
complete TE implementation that support everyone's needs (a:client
b:server c:stub-network d:transit-network). Much of this should probably
become separate standards in the long term (e.g. mechanisms to exchange
locator-selection-data between shim6 hosts in a network). If not, the
basics of shim6 may become way too complex.


//per
-- 
  Per Heldal
  http://heldal.eml.cc/