Re: UDP-Lite as optional transport for CAPWAP
From: Pat Calhoun (pacalhou) (pcalhouncisco.com)
Date: Wed, 14 Mar 2007 15:04:12 -0700 (PDT)
That may be an issue for IPv6, but note that NATs are supposed to go
away with the new address space... I don't have lots of visibility into
this space though.

Pat Calhoun
CTO, Wireless Networking Business Unit
Cisco Systems

 

> -----Original Message-----
> From: Puneet Agarwal [mailto:pagarwal [at] broadcom.com] 
> Sent: Wednesday, March 14, 2007 2:45 PM
> To: Pat Calhoun (pacalhou); David Melman; capwap [at] frascone.com
> Subject: RE: [Capwap] UDP-Lite as optional transport for CAPWAP
> 
> Hi Pat,
> 
> One of the goals of capwap was to go easily through 
> firewalls/nats. Is UDP-lite supported by the existing 
> nat/firewalls (since the nat will have to play around with 
> the checksum which includes the ip psuedo-hdr)?
> 
> Thanks.
> 
> -Puneet
> 
> -----Original Message-----
> From: Pat Calhoun (pacalhou) [mailto:pcalhoun [at] cisco.com]
> Sent: Tuesday, March 13, 2007 12:59 PM
> To: Pat Calhoun (pacalhou); David Melman; capwap [at] frascone.com
> Subject: Re: [Capwap] UDP-Lite as optional transport for CAPWAP
> 
> Let me try again. UDP-Lite should be mandatory when run over 
> IPv6. UDP should continue to be mandatory when run over IPv4.
> 
> Pat Calhoun
> CTO, Wireless Networking Business Unit
> Cisco Systems
> 
>  
> 
> > -----Original Message-----
> > From: Pat Calhoun (pacalhou)
> > Sent: Tuesday, March 13, 2007 7:40 AM
> > To: David Melman; capwap [at] frascone.com
> > Subject: Re: [Capwap] UDP-Lite as optional transport for CAPWAP
> > 
> > I completely agree with this request, and believe that if we want 
> > CAPWAP to work over v6, the change is needed in this version of the 
> > protocol, as requiring the UDP checksum be computed on each 
> IPv6/UDP 
> > CAPWAP packet would have a dramatic performance impact. I do not 
> > believe the feature should be optional for v4 networks.
> > 
> > I have created issue 249. It is currently marked as a wish, 
> but as per
> 
> > my above statement, I believe we should categorize it as a bug.
> > 
> > Pat Calhoun
> > CTO, Wireless Networking Business Unit Cisco Systems
> > 
> >  
> > 
> > > -----Original Message-----
> > > From: David Melman [mailto:davidme [at] marvell.com]
> > > Sent: Tuesday, March 13, 2007 1:29 AM
> > > To: capwap [at] frascone.com
> > > Subject: [Capwap] UDP-Lite as optional transport for CAPWAP
> > > 
> > > Hi.
> > > 
> > > In supporting CAPWAP-over-UDP-IPv6, I'm concerned about 
> the cost & 
> > > performance overhead of supporting a full UDP checksum over
> > the entire
> > > UDP payload.
> > > 
> > > While UDP over IPv4 allows the UDP checksum to be set 0
> > indicating the
> > > UDP checksum is not calculated, UDP over IPv6 requires the
> > UDP header
> > > to contain a checksum which covers the UDP pseudo header
> > and the UDP
> > > payload (per RFC 2460).
> > > Calculating the checksum over the full data payload adds
> > significant
> > > cost and performance overhead, both to the AC switch (typically 
> > > implemented in hardware) and the WTP (typically implemented in 
> > > software).
> > > 
> > > I propose to allow CAPWAP over UDP-Lite (RFC 3828) as an optional 
> > > transport to UDP, particularly for the CAPWAP data channel
> > which would
> > > be impacted the most by the requirement to calculate the
> > UDP checksum
> > > over the entire payload.
> > > 
> > > The UDP-Lite header is identical to the UDP header, except
> > for the UDP
> > > Length field which is replaced with the Checksum Coverage
> > field. The
> > > Checksum Coverage field indicates the number of octets,
> > counting from
> > > the first octet of the UDP-Lite header, that are covered by the 
> > > checksum. To minimize the performance impact, the UDP Checksum 
> > > Coverage is set to the minimal value of 8, which limits the UDP 
> > > checksum to cover just the UDP pseudo-header, excluding the UDP 
> > > payload.  (Note that the UDP Length field is not required,
> > as it can
> > > easily be calculated by subtracting 8 (UDP-Lite header
> > > length) from the IP Length field.)
> > > 
> > > UDP-Lite (RFC2460) is widely available in IPv6 stacks and
> > in all the
> > > common operating systems we surveyed.
> > > 
> > > I believe this change would have minimal impact to the
> > CAPWAP protocol
> > > and has significant value to both hardware and software 
> > > implementations.
> > > 
> > > Regards,
> > > David
> > >  
> > > __________________________________________
> > > 
> > > David Melman
> > > Architecture group - Marvell
> > > Email - DavidMe [at] marvell.com
> > > 
> > > _________________________________________________________________
> > > To unsubscribe or modify your subscription options, please visit:
> > > http://lists.frascone.com/mailman/listinfo/capwap
> > > 
> > > Archives: http://lists.frascone.com/pipermail/capwap
> > > 
> > _________________________________________________________________
> > To unsubscribe or modify your subscription options, please visit:
> > http://lists.frascone.com/mailman/listinfo/capwap
> > 
> > Archives: http://lists.frascone.com/pipermail/capwap
> > 
> _________________________________________________________________
> To unsubscribe or modify your subscription options, please visit:
> http://lists.frascone.com/mailman/listinfo/capwap
> 
> Archives: http://lists.frascone.com/pipermail/capwap
> 

Results generated by Tiger Technologies using MHonArc.