>How common is it for a single server to be *directly* connected to more
>than one customer site? Usually, sites are connected via routers, either
>indirectly through ISPs or directly through "back-door" private links.
Well first we are not sure yet how they will define site? If they use
site as "department" is how I view it and maybe a good assumption for
this discussion as a model to determine the pain.
But DNS and DHCP for sure. Other less known apps are the ones developed
by Oracle, Sybase, Informix etc... where IP addresses are used to
identify clients records (much like DNS and DHCP do).
But once could have:
fec0::1 site A
|
Router
|
fec0::1 site B
Where all DHCP and DNS and Database accesses for both sites
(departments) are in site B. Or department B maintains all servers for
all other departments (other sites).
The problem is that the src address to site B will be feco::1 or from
site A. The server apps needs to know the zone via the sin6_scope_id
from the packet. Then if the database lookup was:
IPv6 address
It would have to change that to:
IPv6 Address + sin6_scope_id
Site B would know this from the router connection to Site A (likewise
Site A would know B).
It really is putting a completely new variable in the net communications
paradigm for these apps.
Bottom line I would like to see us avoid that architecturally for
applications for Ipv6 and keep it simple as it was for IPv4.
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]
--------------------------------------------------------------------
