Re: [Netconf] netconf-binary-encoding comments

[Qin Wu <[e-mail]>]

发件人: Netconf [mailto:[e-mail]] 代表 Mahesh Jethanandani
发送时间: 2018年7月10日 8:06
收件人: Balazs Lengyel
抄送: [e-mail]
主题: Re: [Netconf] netconf-binary-encoding comments




On Jul 9, 2018, at 4:28 AM, Balazs Lengyel <[e-mail]<mailto:[e-mail]>> wrote:

Efficiency is mostly interesting for yang-push data and not so much for <edit-config>  <edit-data> <get-config> <get-data>.
I would generally agree that it is not as interesting for <edit-config> or <get-config>. But we have seen instances of <get-data> where the data set being returned can be rather large, e.g. BGP route entries, where encoding of the data would have helped.

[Qin]: Another complimentary way to improve NETCONF efficiency is to fragment large size data in the RPC message into data chunk in multiple RPC messages, we see this proposal being cooked for quite a long time,
Surprisingly not see any progress on it.

But the question is, do we want to be selective in what we encode? Would it not be easier/simpler for all transaction to be in a single form of encoding?

However we are more interested in dynamic-subscriptions then in configured subscriptions.
Eegards Balazs

On 7/5/2018 7:06 PM, Kent Watsen wrote:

> When I brought up this issue 5 years ago there was zero interest in improving
> the efficiency of NETCONF. Maybe YANG Push will change that POV.

I agree that there is little appetite to improve the efficiency of configuration-oriented
workflows.   Monitoring, for which notifications supports in a big way, have a strong
push for efficiency.  I think the primary question is if this efficiency must be realizable
for NC/RC based subscriptions, or if the efficiency can come from the use of more
suitable transports (e.g., gRPC), that can be defined by future "notif" drafts?

If we were only interested in such efficiency for configured subscriptions, then I think
a "notif" draft to define the transport would be relatively easy.   If such efficiency is
also needed for dynamic subscriptions, then it seems like something along the lines
of what binary-encoding draft proposes might be needed.

Is there a need for efficiency for any other workflow?   What about for RPCs that are
executed often (e.g., polling) or return very large responses?   Does we care about
making these use cases more efficient too, or are we primarily only interested in
just making notifications efficient?

Kent



On 7/5/18, 12:07 PM, "Netconf on behalf of Andy Bierman" <[e-mail]<mailto:[e-mail]> on behalf [e-mail]<mailto:[e-mail]>> wrote:

Hi,

I think the first-order issue is for the WG to decide if there is a problem to be
solved by this WG, and if so, what is the problem scope.

Details like the SID assignments for rpc, rpc-reply, etc do not really impact that decision.

When I brought up this issue 5 years ago there was zero interest in improving the
efficiency of NETCONF. Maybe YANG Push will change that POV.

https://tools.ietf.org/html/draft-bierman-netconf-efficiency-extensions-00<https://urldefense.proofpoint.com/v2/url?u=https-3A__tools.ietf.org_html_draft-2Dbierman-2Dnetconf-2Defficiency-2Dextensions-2D00&d=DwMFaQ&c=HAkYuh63rsuhr6Scbfh0UjBXeMK-ndb3voDTXcWzoCI&r=9zkP0xnJUvZGJ9EPoOH7Yhqn2gsBYaGTvjISlaJdcZo&m=oAFOh8ZwuiodqCiPKKq5-n9X4-VYRoFJXIfDMi8vvJs&s=rPGpbkoE7L63ymXNzHmAKazwljYuJVs3pa2Pf1arT-A&e=>

I think this draft should focus on the protocol mechanisms and not define
any media types. Definitions of GPB and other formats are not trivial and need
their own RFCs.


Andy


On Thu, Jul 5, 2018 at 8:07 AM, Balazs Lengyel <[e-mail]<mailto:[e-mail]>> wrote:
Hello,
Is this applicable for Restconf? Would a Restconf binary encoding be interesting?

Chapter 2)

- A more detailed explanation of which parts are encoded would be good. What is the top XML element that will be encoded? <rpc>, <RPC-reply>, <RPC-error>, <notification> ? Just referencing a figure in another draft is not enough.

- SHOULD, SHALL or SHALL NOT a client server declare support for the XML encoding? Is that always implicit? State it.

4.2) Shouldn't we also have a JSON encoding here?

I would think that all encodings need some official reference, defining how they are used with YANG: RFC, web link

Is the Thrift and gpb encoding trivial or is it described somewhere or do we need an RFC about it? Please state whichever is the case.

regards Balazs





--
Balazs Lengyel                       Ericsson Hungary Ltd.
Senior Specialist
Mobile: [phone]              email: [e-mail]<mailto:[e-mail]>

_______________________________________________
Netconf mailing list
[e-mail]<mailto:[e-mail]>
https://www.ietf.org/mailman/listinfo/netconf<https://urldefense.proofpoint.com/v2/url?u=https-3A__www.ietf.org_mailman_listinfo_netconf&d=DwMFaQ&c=HAkYuh63rsuhr6Scbfh0UjBXeMK-ndb3voDTXcWzoCI&r=9zkP0xnJUvZGJ9EPoOH7Yhqn2gsBYaGTvjISlaJdcZo&m=oAFOh8ZwuiodqCiPKKq5-n9X4-VYRoFJXIfDMi8vvJs&s=g9t4wEhl67_O_ZmVRrFI1qv3g4UdY2tUiwYOV3Xny0o&e=>




--

Balazs Lengyel                       Ericsson Hungary Ltd.

Senior Specialist

Mobile: [phone]              email: [e-mail]<mailto:[e-mail]>
_______________________________________________
Netconf mailing list
[e-mail]<mailto:[e-mail]>
https://www.ietf.org/mailman/listinfo/netconf

Mahesh Jethanandani
[e-mail]<mailto:[e-mail]>