Re: [RTG-DIR] RtgDir review: draft-ietf-rtgwg-policy-model-16

["Acee Lindem (acee)" <[e-mail]>]

Hi Tom, et al, 
In the latest version, -17, we have used the standard IETF ip-prefix type and specified the constrain in the text. 
Thanks,
Acee

On 7/8/20, 7:53 AM, "tom petch" <[e-mail]> wrote:

    From: rtgwg <[e-mail]> on behalf of Yingzhen Qu <[e-mail]>
    Sent: 07 July 2020 19:45

    Hi all,

    Thank you all for the discussions and comments.

    If we’re to do error-checking in the YANG module, here’re the two issues:

    Ip-prefix is a union of “ipv4-prefix” and “ipv6-prefix”. The mask-length-upper is 32 for IPv4 and 128 for IPv6, so we either have to separate ipv4 and ipv6 in order to add a constraint or it will be up to the server to reject the config.

    “ipv4-prefix” and “ipv6-prefix” are defined as a string, so the mask-length is part of the string. I don’t know an easy way to add a constraint if mask-length-lower needs to be verified against mask-length.

    Any comments and suggestions are welcome.

    <tp> 
    As I said before, it is possible to make more checks with YANG but as you point out, they are complex.  For me, it is not worth the added complexity and I would just make sure that the description is clear to such as RTG-DIR.  'must' be present, and max>= min I think fine, but that is as far as I would go.

    Where ipv4 goes one way and ipv6 goes another then YANG choice case case comes to mind but probably not worth it here.

    Tom Petch

    Thanks,
    Yingzhen

    From: John Scudder <[e-mail]>
    Date: Tuesday, July 7, 2020 at 9:06 AM
    To: "John G. Scudder" <[e-mail]>
    Cc: "Acee Lindem (acee)" <acee=[e-mail]>, Yingzhen Qu <[e-mail]>, "[e-mail]" <[e-mail]>, "[e-mail]" <[e-mail]>, "[e-mail]" <[e-mail]>, RTGWG <[e-mail]>
    Subject: Re: [RTG-DIR] RtgDir review: draft-ietf-rtgwg-policy-model-16

    On Jul 7, 2020, at 11:35 AM, John G. Scudder <[e-mail]> wrote:

    Hi Acee,


    On Jul 7, 2020, at 11:16 AM, Acee Lindem (acee) <acee=[e-mail]> wrote:
    Yes. I’d say we should just use the ip-prefix type from RFC 6021. This type has the right semantics.
    However, I’m wondering how we do the mask-length-lower checking with the union. I imagine it should be possible.

    How would you test the constraint on mask-length-upper if you use ip-prefix?

    Duh, that’s what you said, s/upper/lower/. I don’t see a way to do it, but if you can work it out I agree it otherwise seems like the right approach.

    —John