| Re: Comments on -02 version of capwap spec | <– Date –> <– Thread –> |
|
From: Abhijit Choudhury (Abhijit |
|
| Date: Sun, 17 Sep 2006 04:54:59 -0700 (PDT) | |
Title: Re: [Capwap] Comments on -02 version of capwap spec
[Puneet]: My main concern is
the way 802.11 Management traffic
is handled today. It wasn't clear to me why we (the group) specified to
carry them in CAPWAP Data frames. I would have thought that they would
be carried in CAPWAP Control Frames as they would most likely be handled
by the CAPWAP Control entity (and not necessarily the CAPWAP Data
entity). Hence sending such packets to the CAPWAP Data entity seems like
a waste and it would eventually need to be sent to the CAPWAP control
entity.
<PRC> I think architecturally, I disagree. I believe that the CAPWAP
Control entity handles the base spec - maintaining that connection and
configuration with the peer. The CAPWAP Data entity, on the other hand,
is the one that actually understands the subtleties of 802.11 (or
whatever wireless binding). It handles the user associations, the 802.1x
state machine, etc. Further, as I noted above, it becomes complex to
decide what's management and what's not. Clearly, an 802.11 MAC
management frame would be a candidate, but what about EAPOL? What about
IAPP? This would require use to do an extensive protocol survey and list
the proper destination for each.
is handled today. It wasn't clear to me why we (the group) specified to
carry them in CAPWAP Data frames. I would have thought that they would
be carried in CAPWAP Control Frames as they would most likely be handled
by the CAPWAP Control entity (and not necessarily the CAPWAP Data
entity). Hence sending such packets to the CAPWAP Data entity seems like
a waste and it would eventually need to be sent to the CAPWAP control
entity.
<PRC> I think architecturally, I disagree. I believe that the CAPWAP
Control entity handles the base spec - maintaining that connection and
configuration with the peer. The CAPWAP Data entity, on the other hand,
is the one that actually understands the subtleties of 802.11 (or
whatever wireless binding). It handles the user associations, the 802.1x
state machine, etc. Further, as I noted above, it becomes complex to
decide what's management and what's not. Clearly, an 802.11 MAC
management frame would be a candidate, but what about EAPOL? What about
IAPP? This would require use to do an extensive protocol survey and list
the proper destination for each.
I agree with Pat here. The current architecture hands it
cleanly. Sending 802.11 management traffic on the
control
channel
would cause lots of problems unless the 802.1X
and 4-way
handshake were also carried on the control
channel.
Since the latter
are carried in 802.11 data packets, they
definitely belong
on the data channel.
I think the real problem is that the 802.3 tunneling
mode
has not been fully specified. It's not clear how
the
802.11 management frames will be carried in the data
channel
in 802.3 tunneling mode.
Thanks,
Abhijit
-
Re: Comments on -02 version of capwap spec Pat Calhoun (pacalhou), September 5 2006
- Re: Comments on -02 version of capwap spec Puneet Agarwal, September 11 2006
-
Re: Comments on -02 version of capwap spec Pat Calhoun (pacalhou), September 16 2006
- Re: Comments on -02 version of capwap spec Abhijit Choudhury, September 17 2006
- Data and Control paths, was: Re: Comments on -02 version of capwap spec David T. Perkins, September 17 2006
- Re: Comments on -02 version of capwap spec Puneet Agarwal, September 20 2006
- Re: Comments on -02 version of capwap spec Pat Calhoun (pacalhou), September 17 2006
- Re: Comments on -02 version of capwap spec Pat Calhoun (pacalhou), September 20 2006
Results generated by Tiger Technologies using MHonArc.