Keith;
I disagree with your assessment. I will continue this technical discussion
on the WG list after the minutes are published.
Rich
At 06:26 PM 3/27/03 -0500, Keith Moore wrote:
> I second Tony's key point. SL's are just 1 form of IPv6 addresses
> with a limited scope. As soon as operations folks put up firewall or
> router filters, global addresses have the same scope limitations.
they don't have the same set of problems that SLs do.
SL addresses are ambiguous. if you can't reach a host using its SL
address, you don't know whether the problem is that you're in a
different site or whether the host is down or whether there is a link
failure or this is prohibited by policy. so a multiparty app ends up
needing to implement various hacks to deal with ambiguous addresses
(proxies, tunneling, etc), in order to function across site boundaries
(and apps *will* be expected to function across site boundaries).
with globals, if an app can't reach the host using a global address,
it's either a host failure, a network failure, or a policy decision. to
a large degree this can be disambiguated using ICMP. but in many cases
the app can treat these as 'out of its control' since there's no way to
work around them.
SLs thus break a clean separation of function between the apps and the
network.
Keith
------------------------------------
Richard A. Carlson e-mail:
RACarlson(_at_)anl(_dot_)gov
Network Research Section phone: (630) 252-7289
Argonne National Laboratory fax: (630) 252-4021
9700 Cass Ave. S.
Argonne, IL 60439