Le 05/09/2014 14:45, Alper Yegin a écrit :
Hi Alex,
On Sep 5, 2014, at 3:32 PM, Alexandru Petrescu wrote:
Le 05/09/2014 10:48, Alper Yegin a écrit :
Alex,
DMM is not meant to be only about a bunch of MIP-based
solutions. There are various components in DMM solution space
that'd also work with GTP-based architectures. For example,
identifying the mobility needs of flows. Or, conveying the
mobility characteristic of a prefix to the UE.
Alper - thanks for the reply.
Identifying the mobility needs of flows assumes that IP flows _can_
be characterized, and then distinguished as mobile - or non-mobile.
I think this is very hard to do, given the difficulty to write good
firewall rules, and the difficulty of analyzing traffic dumps.
For example, Netalyzr was written to tell whether or not one's
computer is connected to the Internet. That report page has so
many lines that it is hard to tell which part of it really means
'connected to the Internet'.
The same problem may arise when trying to identify a particular
'flow'.
The most robust way is to let the application tell the IP stack.
http://www.ietf.org/id/draft-yegin-dmm-ondemand-mobility-02.txt
Sounds reasonable.
A complimentary means is to look at this as a source address selection
problem: given two addresses configured on an interface (an IP address,
and an IP address designated as CoA), the application selects the src
address as the IP address if its flows are brief query/response (http),
or it selects the src address as the CoA if its flows are longer timed
(request of UDP stream, or TCP download).
This would need a means to designate to the stack an IP address as to be
a 'CoA', or otherwise designate a prefix to be the 'home' prefix and, by
deduction any address differing be the CoA.
Alex
Alper
Alex
Alper
On Sep 4, 2014, at 1:14 PM, Alexandru Petrescu wrote:
Le 03/09/2014 20:53, Brian Haberman a écrit :
Behcet,
On 9/3/14 2:33 PM, Behcet Sarikaya wrote:
You don't seem to understand my points.
That is quite possible. Your comment on the list was "I am
against any deployment work before we decide on a
solution..."
I read that as an objection to having the deployment models
work item on the agenda. Please do tell me what I am
missing.
Regards, Brian
Hi,
I am following the discussion and me too I do not quite
understand what is the complain.
I am happy to learn that a if a WG is to be formed then it
would be around a solution rather than just requirements or
architecture.
That said, I would like to express a worry along similar
lines.
In DMM, precedents and the keen NETEXT, there seems to be a
hard-rooted disconnect between the product developped -
(P)Mobile IP - and the deployments. We know for a fact that
3GPP deployments (2G/3G/4G) do not use (P)Mobile IP. We also
know that 3GPP specs do mention Mobile IP. To such a point that
I wonder whether 3GPP has not the same disconnect as here.
On another hand, we do have indications of where (P)Mobile IP
is used - the trials, the projects, the kernel code, and not
least the slideware attracting real customers.
The worry: develop DMM protocol while continuing the
disconnect.
Alex
_______________________________________________ dmm mailing
list [email protected] https://www.ietf.org/mailman/listinfo/dmm
_______________________________________________ dmm mailing
list [email protected] https://www.ietf.org/mailman/listinfo/dmm
_______________________________________________
dmm mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/dmm
