On Thu, Feb 28, 2002 at 01:41:09PM -0800, Charles E. Perkins wrote:
Actually, we hope to get it to work without requiring X.509.
Heh. I agree with you, I threw in X.509 just to make the problem seem
much more intractable. :-)
I can think of making it the mobile IP problem more soluble, such as
informing an agent at my local ISP (with whom I have a trust
relationship), and if my local ISP then had trust relationships with
its routing peers, it could pass the message along, so that it would
vouch for the redirect request. That solves the authorization problem
by reducing it to a previously solved problem.
But then this makes it like multicast; I still won't be able to get
the benefits of Mobile IP unless not only a sigificant portion of the
routing core adopts it, but my local ISP is going to have to adopt it
is as well. And if the deployment experience for mobile IP is
anything like multicast, I'm not going to hold my breath waiting for
my ISP to offer it to me as a service.
In contrast, DDNS is something I could set up on my own, since I
control my own domain server. (Yeah, I know not everyone has this
luxury, but I strongly suspect it will be a lot easier to get your DNS
service handled by someone other than your ISP.)
The approach you favor would require resolution via DNS after
every movement. That's going to be a disaster for smooth handovers,
I reckon.
It does require short TTL's on the DNS record, and it does assume that
in general there aren't long-lived TCP connections that need to
survive a rehoming operation. On the other hand, for most operations
initiated on mobile hosts (i.e., initated TCP connections to the http,
https, imap, smtp, etc. ports), the other side will very likely not
care at all about the originating IP address, and no DNS resolution
would be required.
I can certainly think of applications for which the handoffs wouldn't
be very smooth --- for example, if my laptop where an web server for a
very busy and frequently visited web site, and I wanted people to find
me whereever I set up my laptop, whether it's at a NordU 2002
conference in Finland or an IETF terminal room in Yokohama. But is
that really a realistic scenario? There will already be extremely
non-smooth handoffs caused by the fact that my laptop won't be on the
internet while it's travelling over the Atlantic or Pacific oceans.
So for all of the applications I can think of today, where devices are
mobile, smooth handoffs and support for TCP connections that last
longer than a longer than a single terminal room session aren't what
I'd call even a remotely common case. Maybe things will be different
in the future; but will there be enough mobile devices which will be
hosting high-traffic services, or which need to support very
long-lived TCP connections? I don't see it, myself.
- Ted