RE: [Fwd: I-D ACTION:draft-lear-multi6-things-to-think-about-00.txt]

["Tony Li" <Tony.Li@procket.com>]

Brian,

I thought that the point of this document was things to think about,
not a set of decision criteria.  If it is a set of criteria, then aren't
we duplicating the requirements draft?  No, I don't want to reopen that,
but what I *AM* suggesting is that Eliot's list is literally a list
of discussion questions.  As such, I think that this is a fine question
to ask.

If you're elevating this draft as anything other than a set of 
discussion questions, then we need to have a different discussion.

Which is it?

Tony


|  -----Original Message-----
|  From: Brian E Carpenter [mailto:brc@zurich.ibm.com] 
|  Sent: Thursday, December 18, 2003 12:54 AM
|  To: Multi6
|  Subject: Re: [Fwd: I-D 
|  ACTION:draft-lear-multi6-things-to-think-about-00.txt]
|  
|  
|  Pekka Savola wrote:
|  > 
|  > On Tue, 16 Dec 2003, Tony Li wrote:
|  > > I'd submit that if it cannot be easily ported back to 
|  IPv4, that it
|  > > bears closer examination.  v4 and v6 are architecturally very
|  > > similar.  Any solution that does not apply to both is either a
|  > > kludge or is exploiting an odd property of one of the two.  In
|  > > either case, it would bear close examination.  You know, the kind
|  > > that you give things when the fire alarm in the building 
|  goes off...
|  > > ?  ;-)
|  > 
|  > I have to heartily disagree here.  IPv6 address *does* 
|  have more bits.
|  > Different problem spaces have leveraged that property 
|  before as well,
|  > leading to solutions which are not easily backportable to IPv4.
|  > 
|  > Maybe one could reword this differently: the solution beas some
|  > thinking about if it doesn't rely on the 128bit address length of
|  > IPv6, and is not easily IPv4-capable.
|  
|  Yes. But this is the IPv6 multihoming WG, so while applicability to
|  IPv4 is an interesting question to ask, it cannot be a decision
|  criterion.
|  
|  I would counter-argue against Tony in another way. If we had variable
|  length addresses (as some people strongly suggested for IPng) a whole
|  new class of multihoming solutions might be available. But we don't,
|  so they aren't. Thus, you cannot argue that solutions *must* 
|  be independent 
|  of address length considerations.
|  
|  (Please don't kick off a thread on variable length addresses... at
|  least not here.)
|  
|    Brian
|  
|