Rich,
>> I don't agree we can declare this in the spec. How we word
>> it I am not
>> sure yet. But mobile nodes will use global addresses not
>> site-local if
>> they are roaming across routing realms. Site-local simply won't work.
>You are getting at the interaction between site-local addresses and
>mobility. Can a mobile node have a site-local home address as well as a
>global home address? If the implementation does not support mobility for
>scoped addresses, then the destination address selection will need to be
>modified.
I see that too. But I am coming from a much stronger issue. I do
not think the src addr selection work should prefer less scope over
greater scope or equal scope ever as a default.
>How about a rule like this: if you are comparing two destination addresses
>DA and DB, and Source(DA) is a home address and Source(DB) is not a home
>address, then sort DA before DB.
Thats fine but this is getting out of control now we have this to check.
But this sounds right. But I really think if this causes any
performance issues for servers that handle 50,000 requests and many
simultaneously this entire source address selection algorithm/config
will get shut off and rely on DNS ordering which can be done too at
least with BIND DNS.
I strongly suggest we have a common set of sorts that are basic and to
what we have deployed today plus 6t04 and leave the others in the spec
as TBD. No one has shipped a base OS product or router supporting
scoping to the degree we are discussing hear and it has not been tested
at any bakeoffs.
regards,
/jim
--------------------------------------------------------------------
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 [EMAIL PROTECTED]
--------------------------------------------------------------------
