Susan,

Jeff Zhang and I've reviewed info model policy draft.

One general comment (and I'm repeating somebody else). Didn't we decide to use 
YANG for all modeling? I see you use RBNF, so it would be nice to have YANG 
models in the draft.

On Jul 7, 2014, at 7:28 PM, Susan Hares 
<sha...@ndzh.com<mailto:sha...@ndzh.com>> wrote:

My co-authors and I wish feedback on the I2RS informational models:

1)      Draft-hares-i2rs-info-model-policy
Contains: Basic Network Policy IM
                   Policy-Based Routing IM
                   I2RS Local configuration

http://datatracker.ietf.org/doc/draft-hares-i2rs-info-model-policy/


In I2RS Local Config Information Model (I2RS-LC IM) there is a typo

and provides quick local access to polies

/polies/policies

In fig 5, both arrows go from action to condition. I couldn't figure out what 
is the meaning of use two arrows pointing to same direction.

What's the relationship between policy rule and policy? E.g., in the following:

  The elements of the Policy Rule information model are as follows:

  o  A policy can in turn be part of a hierarchy of policies, building
     on top of other policies.  Each policy is captured in its own
     level, distinguished via a policy-identity.

Is the "policy" a "policy rule" or "policy group"?

In the following:

  o  A condition contains a variable and a value and use a match
     operator, to connect variable with value.  An examples of an
     operator might be a" IP ADDRESS AS RESOLVED BYDNS" or "Set to a
     member".

The example is rather confusing for a match operator.

Page 12 typo

IP ADDRESS AS RESOLVED /BYDNS/BY DNS

What's the difference between priority and precedence?

       <Policy-Rule_priority> ::= INTEGER (0..250);
       <Policy-Rule_precedence> :;= INTEGER (0..250);

Policy constraints may be exchanged by routing protocols while updating routing 
information.
This can cause the routing protocols to constantly updated information, as the 
policy might be in direct conflict with routing protocol configuration on 
another device(s). Why do you want to do it? I'm in favor on policies applying 
only on the device where policies are explicitly applied.

Is section 3.5 complete?

For the following:

  o  PRB Default RIB - default forwarding FIB.

Is it talking about FIB or RIB? It's conflicting.


In figure 7 on page 18, there is no relation between condition and rule, as in 
fig 5. Shouldn't in all rules the relation between condition and action be the 
same? If condition met, follow by the action?

On page 22, why is the

Local Policy Information Model abbreviated with LB IM

it got me confused reading below, as I was expecting LP IM.

The PBR IM seems incoherent:

4.2.  PBR-RIB definition

  ...   Each PBR RIB has the following:

  o  PRB RIB NAME
  o  PBR Route-entry

  The Route entry in a PRB has the following information:

  o  match field - as in the RIB IM route

  o  order_list PBR route list with each entry having: a) next-hops, b)
     PBR route attributes, and c) vendor-attributes

  The PRB route attributes include QOS Attributes as show in the policy
  list below.

It talks about route entries in PBR RIBs, with the same match field as in the 
regular RIB IM, i.e. route prefixes. It then talks about each entry has ordered 
list of "PBR route" with next-hops and PRB route and vendor-attributes.

So I assume the PBR route attributes are related to policies. However, the rest 
of the text does not have a coherent connection with that. In fact, the above 
conflicts with section 4.6.

Additionally, is the QOS attributes mentioned above for matching or action? Is 
QoS specific to PBR?

Figure 7 shows QoS action only. It shows "PBR Condition" but that is not 
elaborated.

I can understand that a PBR rule extends a Policy Rule, but I don't understand 
how "QoS action" and "Forward Action" extend "PBR action". Those two seem to be 
part of the "PBR action"?

What does it mean by "Nexthop Type 'extends' Nexthop Variable"?

>From the following in section 4.4:

       <Policy-Rule_Match_Node_PBR-IM> ::= <IPv4_QoS_Node_Matches>
                      | <IPv6_QoS_Node_Matches>

It seems that PBR is all about QoS?

Why do we have the following three things that actually are the same?

          <IPv4_QoS_Node_matches> = <IPv4-QOS_Matches>
          <IPv4_QoS_Value_matches> = <IPv4-QOS_Matches>

What does the following mean?

       <Policy-Rule_Match_Operator_PBR-IM> ::= [<Longest-prefix>]
                  | [<Exact>]
                  | [(<IPv4-RANGE> <IPv4-Low> <IPv4-High>)]
                  | [(<IPv6-RANGE> <IPv6-Low> <IPv6-High>)]
                  | [(<LENGTH-Range> <LENGTH_Low> <LENGTH_High>)]

Jeff and Dean



_______________________________________________
i2rs mailing list
i2rs@ietf.org
https://www.ietf.org/mailman/listinfo/i2rs

Reply via email to