I understand your point below "they were viewed as more general purpose than just
for diffserv." I can see uses with respect to admission control and resource allocation (i.e. intsrv). And you are correct regarding the use of the frwkPrcSupport table for the purpose of the PEP notifying the PDP whether or not it supports 802-based classification. I just do not care for the fact that the PDP would have to parse two EPDs prior to being able to format "Filter" EPDs.
Tina Iliff
MCIWorldCom ENSD
(972)729-1620
-----Original Message-----
From: Hahn, Scott [mailto:scott.hahn@intel.com]
Sent: Wednesday, October 18, 2000 1:18 PM
To: Iliff, Tina; 'rap@ops.ietf.org'
Subject: RE: [Diffserv] DiffServ PIB
I envision that other classifiers will be added in their own PIBs.
Since the 802 (and the IP) classifiers extend the base filter, they are not
required. The existence of the 802 classifier in the framework PIB just
allows any device that can do 802 level classification to use this standard
template for defining the classifier. Early on we decided that we needed to
have some classifiers defined "out of the box" in order to make the PIB
usable. The classifiers we chose were the IP and 802 classifiers because we
believed that they were commonly supported by many network devices. These
classifiers were added to the framework PIB they were originally included in
the Diffserv PIB) because they were viewed as more general purpose than just
for diffserv. Currently the Diffserv PIB only allows L3 and L4 info to be
passed up in the qosIfClassificationCaps, but you can still specify the 802
classifier as a supported PRC and while I haven't looked into it too far, I
believe you could use the compsLimit table to specify restrictions.
If there is a strong feeling that the 802 classifier definition be
moved to it's own PIB, that can be done, though my preference is to leave it
in the framework PIB.
All that said, I don't think that the only classifier useful, at the
edge, the to determine a PHB and assign DSCPs will be the traditional IP
6-tuple. In fact I assume that edge devices will regularly use other
classifiers to determine what DSCP to mark/re-mark packets.
-Scott
-----Original Message-----
From: Iliff, Tina [mailto:Tina.Iliff@WCOM.Com]
Sent: Wednesday, October 18, 2000 7:47 AM
To: 'rap@ops.ietf.org'
Subject: FW: [Diffserv] DiffServ PIB
All,
Please refer to below e-mail traffic...
Maybe the tables in the Framework PIB need to be relooked at....in my
opinion the frwk802Filter extension to the frwkBaseFilterTable is not global
to all COPS-PR clients. The specific filter definitions should be included
in a separately specified PIB. Let me know if I am offbase here...
Thanks,
Tina Iliff
MCIWorldCom ENSD
(972)729-1620
-----Original Message-----
From: Brian E Carpenter [mailto:brian@hursley.ibm.com]
Sent: Wednesday, October 18, 2000 9:27 AM
To: Iliff, Tina
Cc: 'diffserv@ietf.org'; Rawlins, Diana; Sunlin, Richard (c); Bert
Wijnen; Scott Bradner
Subject: Re: [Diffserv] DiffServ PIB
Tina,
In a word, yes. 802 issues are not part of this WG's charter. Either RAP or
ISSLL would need to tackle this one. (I don't mean to suggest it is
unimportant - but WGs need to stick to their charters.)
Brian
> "Iliff, Tina" wrote:
>
> Brian,
>
> Does your reply imply the we ignore the frwk802FilterTable in the
Framework PIB until someone develops Layer 1 and Layer 2 QoS
> PIBs?
>
> Tina Iliff
> MCIWorldCom ENSD
> (972)729-1620
>
> -----Original Message-----
> From: Brian E Carpenter [mailto:brian@hursley.ibm.com]
> Sent: Wednesday, October 18, 2000 3:22 AM
> To: Iliff, Tina
> Cc: 'diffserv@ietf.org'; Rawlins, Diana; Sunlin, Richard (c)
> Subject: Re: [Diffserv] DiffServ PIB
>
> Tina,
>
> In my opinion the diffserv PIB needs to be blind to lower layer
issues,
> i.e. should in no circumstances refer to 802 or any other lower
layer
> technology. Mapping to lower layers should be handled by a
separate
> PIB or MIB.
>
> Brian
>
> > "Iliff, Tina" wrote:
> >
> > All,
> >
> > I have a comment on the DiffServ PIB:
> > Specifically, on pg. 10, the qosIfClassificationCaps attribute
definition:
> >
> > 1. Recommend changing the attribute to
qosIfClassificationCapsDefn or qosIfClassificationCapsSpec
> > 2. The definition is not in line with the
frwkBaseFilterTable and its extends in the Framework PIB I-D; i.e. one
> of the
>
> > extends is the frwk802FilterTable which allows for the
provisioning of 802-based filter configuration data.
> I recommend
>
> > that more bits be added to the SYNTAX definition of the
qosIfClassificationCaps attribute. This will allow
> for PDP
>
> > discrimination in regards to including the 802-based
attributes in the provisioning message instead of
> receiving an error
>
> > from the PEP due to it being unable to support 802-based
filtering. It will minimize bandwidth, connection
> losses due to
>
> > inability to process/corruption due to processing
problems, and configuration time.
> >
> > Tina Iliff
> > MCIWorldCom ENSD
> > (972)729-1620