On Thu, Apr 19, 2018 at 9:54 AM, Robert Varga <[email protected]> wrote:

> On 19/04/18 17:58, Anil Vishnoi wrote:
> >
> >     ... it is matter of what is the correct model. Given that OpenFlow
> has
> >     no notion of a zone, I believe ipv4-address-no-zone is the correct
> type
> >     representing OFPXMT_OFB_IPV4_*.
> >
> > ​ Please do not see it from specific project perspective, neither a code
> > generator should care what is really used and what is not. They go by
> > what is defined in yang model and generate clean interfaces.
>
> Yes and I am talking very specifically about how OFP models (which *are*
> openflow-centric) are using ipv4-address, where I believe they should
> use ipv4-address-no-zone.
>
​Yes agree, but technically existing use of ipv4-address is also not wrong.
So if that's the case, it should not fail.

>   ​
> >     To illustrate, can you explain what
> >     happens when a user specifies two matches, which differ only in zone,
> >     for example "1.2.3.4%a" and "1.2.3.4%b"?
> >
> > ​ OpenFlow plugin ​should fail, if it's not supporting zone, but user
> > can always write an application who uses zone. I think it's none of
> > mdsal concern on how consumer is using these data type, neither mdsal
> > should be influence by the context of consumption model when it comes to
> > implementing defined data types.
>
> Except OFP does not fail, it strips the zone and proceeds merrily, for
> example here:
> https://github.com/opendaylight/openflowplugin/blob/a862a004
> 11d1eadd7db500711723f9bc273580db/openflowplugin/src/main/
> java/org/opendaylight/openflowplugin/openflow/md/core/sal/
> convertor/common/IpConversionUtil.java#L253
> . Wouldn't it be better to explicitly express the fact that OpenFlow
> protocol simply has no notion of a zone?
>
​Yes it would be better, and this is not my question here. For example
assume openflow support zone and they want to use ipv4-address derive type
(which they are using) so ideally it should not fail just because mdsal now
support no-zone? That's most probably because the way MDSAL generating code
(probably rightly! as per it's current design).

>
> And no, MD-SAL is not taking sides, I am not sure where you inferred
> that from.
>
​That's because you are talking more about openflow should use no-zone,
because that's the right way to do for openflow, which i agree with. But my
question here is about the derive types, if ipv4-address-no-zone and
ipv4-address are two derive types and user are free to use any of these
derive type. Currently openflowplugin is using ipv4-address, so why they
have to change their yang model to use no-zone.
https://git.opendaylight.org/gerrit/#/c/71083/3/extension/
openflowjava-extension-nicira/src/main/yang/nicira-action.yang

[#]https://git.opendaylight.org/gerrit/#/c/71083/

>
> Regards,
> Robert
>
>


-- 
Thanks
Anil
_______________________________________________
openflowplugin-dev mailing list
[email protected]
https://lists.opendaylight.org/mailman/listinfo/openflowplugin-dev

Reply via email to