+1

It should represent business logic rather than a  subset of existing features.

Regards,
Jeff

> On Aug 23, 2016, at 2:41 PM, Sam Aldrin <[email protected]> wrote:
> 
> Model design shouldn't be limited by the device capabilities, rather it 
> should be agnostic.
> The existing IETF model is more of a YANG version of CLI, which is rather 
> limiting from operational perspective.
> 
> How do we (operators) like to see ACL model should be? take a look at the 
> model we published (work in progress) at 
> <https://github.com/openconfig/public/tree/master/release/models/acl>
> 
> -sam
> 
>> On Mon, Aug 22, 2016 at 7:04 AM, Adrian Pan <[email protected]> wrote:
>> Hi Dean,
>> 
>>  
>> 
>> 3)   With the model definition, even the acl-type is configured as Ethernet, 
>> the operator still can configure the matches of ace under the acl as ipv4 or 
>> ipv6, right?
>> 
>>  
>> 
>> No, if ACL type is ethernet, then all ACEs are expected to be ethernet. 
>> 
>> [Adrian] I understand your point, but this is not reflected in the model, if 
>> according to the model, the operator still can configure the acl-type as 
>> Ethernet, while configure the ace of the acl as ipv4, and this should be 
>> valid configuration.
>> 
>> 
>> is this the model design intention?
>> 
>>  
>> 
>> If acl-type is of one family, then only ace with match condition from that 
>> family are expected to be in the acl. If you want to combine them, please 
>> use mixed type.
>> 
>> [Adrian] if it’s only expected to be the same as the acl-type, but without 
>> the restriction in the model, you can’t avoid the operator configuration to 
>> mix the acl-type and the ace matches. So my thinking is that, can we add the 
>> restriction in the model for this as below to better reflect the model 
>> design intention?
>> 
>>  
>> 
>>  
>> 
>>  
>> 
>> container matches {
>> 
>>   description
>> 
>>     "Definitions for match criteria for this Access List
>> 
>> Entry.";
>> 
>>  
>> 
>>   container ace-ipv4 {
>> 
>>     when "../../acl-type='ipv4-acl'";
>> 
>>     description "IPv4 Access List Entry.";
>> 
>>     uses packet-fields:acl-ip-header-fields;
>> 
>>     uses packet-fields:acl-ipv4-header-fields;
>> 
>>   }
>> 
>>   container ace-ipv6 {
>> 
>>     when "../../acl-type='ipv6-acl'";
>> 
>>     description "IPv6 Access List Entry.";
>> 
>>     uses packet-fields:acl-ip-header-fields;
>> 
>>     uses packet-fields:acl-ipv6-header-fields;
>> 
>>   }
>> 
>>   container ace-eth {
>> 
>>     when "../../acl-type='eth-acl'";
>> 
>>     description
>> 
>>       "Ethernet Access List entry.";
>> 
>>     uses packet-fields:acl-eth-header-fields;
>> 
>>   }
>> 
>> }
>> 
>>  
>> 
>>  
>> 
>> Thanks
>> 
>> Adrian
>> 
>>  
>> 
>> From: Dean Bogdanovic [mailto:[email protected]] 
>> Sent: Monday, August 22, 2016 5:39 PM
>> To: Adrian Pan <[email protected]>
>> Cc: [email protected]; [email protected]; [email protected]; netmod WG 
>> <[email protected]>
>> Subject: Re: Question about acl-type in draft-ietf-netmod-acl-model-08
>> 
>>  
>> 
>> (+netmod mailing list)
>> 
>> Adrian,
>> 
>>  
>> 
>> Please see inline
>> 
>> On Aug 22, 2016, at 2:27 AM, Adrian Pan <[email protected]> wrote:
>> 
>>  
>> 
>> Dear authors,
>> 
>>  
>> 
>> I have some questions about ietf acl model as below, your reply is 
>> appreciated.
>> 
>>  
>> 
>> 1)   In the model definition acl-type is one key of the acl, also in the 
>> description it says that the acl-type could be ethernet, IPv4, IPv6, mixed, 
>> in case the acl-type is mixed, what’s the identifier should be?
>> 
>> Should it be augmented by different vendor? Since I don’t  see the 
>> definition about it.
>> 
>>  
>> 
>> As mixed ACLs are not supported by all vendors, those are not part of the 
>> standard model. Iit is up to the vendor to augment the ace-type and select 
>> an identifier to their liking. 
>> 
>> 
>> 
>> 
>> 2)   In the “mix” case, the “matches” the ace list can be the combination of 
>> Ethernet,ipv4,ipv6 for different ace, right?
>> 
>>  
>> 
>> Or another combination, again depends on what that particular vendor 
>> supports.
>> 
>> 
>> 3)   With the model definition, even the acl-type is configured as Ethernet, 
>> the operator still can configure the matches of ace under the acl as ipv4 or 
>> ipv6, right?
>> 
>>  
>> 
>> No, if ACL type is ethernet, then all ACEs are expected to be ethernet. 
>> 
>> 
>> is this the model design intention?
>> 
>>  
>> 
>> If acl-type is of one family, then only ace with match condition from that 
>> family are expected to be in the acl. If you want to combine them, please 
>> use mixed type.
>> 
>>  
>> 
>> Dean
>> 
>> 
>> 
>> 
>> module: ietf-access-control-list
>>    +--rw access-lists
>>       +--rw acl* [acl-type acl-name]
>>          +--rw acl-name               string
>>          +--rw acl-type               acl-type
>>          +--ro acl-oper-data
>>          +--rw access-list-entries
>>             +--rw ace* [rule-name]
>>                +--rw rule-name        string
>>                +--rw matches
>>                |  +--rw (ace-type)?
>>  
>> 
>>          leaf acl-type {
>> 
>>            type acl-type;
>> 
>>            description
>> 
>>          "Type of access control list. Indicates the primary intended
>> 
>>          type of match criteria (e.g. ethernet, IPv4, IPv6, mixed, etc)
>> 
>>          used in the list instance.";
>> 
>>          }
>> 
>>  
>> 
>> 
>> 
>> 
>> Thanks
>> 
>> Adrian
>> 
>>  
>> 
>> 
>> _______________________________________________
>> netmod mailing list
>> [email protected]
>> https://www.ietf.org/mailman/listinfo/netmod
> 
> _______________________________________________
> netmod mailing list
> [email protected]
> https://www.ietf.org/mailman/listinfo/netmod
_______________________________________________
netmod mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/netmod

Reply via email to