[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: Draft on Globally Unique IPv6 Local Unicast Addresses



kre,

I won't disagree. I just don't want people to believe that
we can define an addressing format, a scope rule, and we're done.

   Brian

Robert Elz wrote:
> 
>     Date:        Thu, 15 May 2003 13:41:03 +0200
>     From:        Brian E Carpenter <brian@hursley.ibm.com>
>     Message-ID:  <3EC37CCF.3F753DAC@hursley.ibm.com>
> 
>   | The trouble is that none of this answers the real world questions
>   | that a sophisticated applications environment ought to want to ask,
>   | such as:
> 
>   | I'm increasingly concerned that there is no way to resolve such questions
>   | by examining the prefix bits in an address or by defining scope as a set
>   | of concentric circles such as link-site-global.
> 
> Why is this a surprise?
> 
> Of course answers to those questions aren't going to be as simple as
> looking at a few bits - we shouldn't expect them to be, nor should we
> be criticising the bits for being unable to provide those answers.
> 
> This is obvious as soon as you notice that the answers to the questions
> have a temporal component - which addressing manages to select the
> link which is most reliable (etc) varies as link conditions change, yet
> we obviously don't want the addressing to do so.
> 
> Network boundaries could be delimited by addressing - but only as long
> as the network boundaries are nice fixed stable things, planned well
> in advance.   That certainly won't serve every need.
> 
> But that we cannot possibly solve every possible problem that exists
> with addressing doesn't mean that we shouldn't get whatever advantages
> that we can.   We need stable local addressing.  Whether that is the SL
> addresses that we have now, or the addresses in Bob's proposal (which
> despite his desire to make them different from SL's, have ended up
> being essentially the same - currently the one big difference is the
> bigger address space, that's really it, and that there are no multi-site
> nodes in his proposal).    That this happened is no surprise - there's
> a demand that this kind of addressing fills, and everyone knows it.
> 
> kre
--------------------------------------------------------------------
IETF IPng Working Group Mailing List
IPng Home Page:                      http://playground.sun.com/ipng
FTP archive:                      ftp://playground.sun.com/pub/ipng
Direct all administrative requests to majordomo@sunroof.eng.sun.com
--------------------------------------------------------------------