Re: [Isis-wg] draft-ietf-isis-yang-isis-cfg-07

["Les Ginsberg (ginsberg)" <[e-mail]>]

Radha -

You are making the mistake of thinking that vendor specific CLI determines how we define the YANG model.

The protocol functionality is:

1)Define system-id for an IS-IS instance
2)Define one or more area addresses for the IS-IS instance

Whether a particular vendor uses a "net" config or  a "system-id" and "area address" config isn't relevant.

   Les

(Of course I work for a vendor that uses "net" config. :) )


From: Radha Danda [mailto:[e-mail]]
Sent: Wednesday, November 25, 2015 8:47 AM
To: Les Ginsberg (ginsberg); [e-mail]
Subject: RE: draft-ietf-isis-yang-isis-cfg-07

Thanks Les.

Regarding the net-address configuration. Find the sample configuration from Cisco router
router isis
net 49.0001.1720.1600.1001.00
net 50.0001.1720.1600.1001.00

net-address here consists of area-address, system-id and selector. Even Juniper router has similar configuration. So I think yang should be modelled to support these existing configurations.
And validation should be done to ensure the uniqueness of system-id.

In the above mentioned net configurations, system-id is unique.

If we don't support net-address configuration in yang, I am afraid back-end implementation of protocols might have to be changed. Please let me know your thoughts.

Also the way yang model is currently designed (assuming the selector is removed from system-id), there is no way to input selector now.

And separate area-address is not needed if we support net-address configuration.

Best Regards,
Radha

From: Les Ginsberg (ginsberg) [mailto:[e-mail]]
Sent: Wednesday, November 25, 2015 9:33 PM
To: Radha Danda; [e-mail]<mailto:[e-mail]>
Subject: RE: draft-ietf-isis-yang-isis-cfg-07

Radha -

From: Isis-wg [mailto:[e-mail]] On Behalf Of Radha Danda
Sent: Wednesday, November 25, 2015 7:49 AM
To: [e-mail]<mailto:[e-mail]>
Subject: [Isis-wg] draft-ietf-isis-yang-isis-cfg-07


Hi,

I was going through draft-ietf-isis-yang-isis-cfg-07 and I have below comments.



1. system-id definition should not contain selector. As per NSAP definition, system-ID is 6 bytes and it has below mentioned format and it is only of length 6 bytes.





[Les:] I agree - selector should NOT be part of system-id.



typedef system-id {

    type string {

      pattern

       '[0-9A-Fa-f]{4}\.[0-9A-Fa-f]{4}\.[0-9A-Fa-f]{4}';

    }

    description

     "This type defines ISIS system id using pattern,

      system id looks like : 0143.0438.AeF0";

  }



2. Also we need support to configure net-address as shown below.



[Les:] No. The model correct supports the config of one system-id (as modified above) and one or more area addresses.

There are implementations which combine the configuration into a single CLI "net" - but if multiple area addresses are supported the same system-id MUST be entered on each line - so functionally what you have are multiple (synonymous) area addresses and a single system-id. The model has it correct in that regard.



   Les



       typedef net-address {

                type string {

                        pattern

                                '[0-9A-Fa-f]{2}\.([0-9A-Fa-f]{4}\.){0,3}[0-9A-Fa-f]{4}\.[0-9A-Fa-f]{4}\.[0-9A-Fa-f]{4}\.00';

                        configd:pattern-help "XX.XXXX. ... .XXXX.XX";

                        configd:help "Network entity title (NET)";

                }

        }



Please let me know if you need more details. This is my first post, so please excuse me in case if I miss to provide some details.



Best Regards,

Radha