Thanks Jeff for review + detailed comments, please see inline.

Co-authors, feel free to chime in. We¹ll meet to discuss in Chicago.


On 2017-03-24, 12:14 PM, "Jeffrey Haas" <[email protected]> wrote:

>Authors,
>
>Thanks for continued work on the yang module.  It does appear it's
>approaching a good level of maturity!
>
>I have a few comments from my recent read-through.  I'd like to encourage
>other members of the Working Group to review the module.  I'd also like to
>encourage the authors to reach out to other protocol groups to check the
>suitability of the module for interaction with their protocols.
>
>Some stream of thought comments:
>- TTL seems to be only a property of multi-hop sessions.  How do you
>  configure GTSM for single-hop sessions?

<RR> TTL must be 255 as per section 5 of RFC5881?
 
>- In single-hop, sessions are operationally keyed on "interface".
>However,
>  there's also a "out-interface" node.  How are they intended to be
>  different?

<RR> This is for the case where ³interface² is a virtual interface,
³out-interface² would indicate the physical interface. Thinking more about
this, we should consider taking it out of the base model, vendors who want
to support this can do augment.

>- state-change-reason in notification is a string type.  While this is
>  probably the fastest path to getting some code written, it's also the
>  least machine usable mechanism.  Has there been any discussion among the
>  authors on how that might otherwise be addressed?

<RR> You¹re the first one to bring it up :-) So you¹d like us to e.g. look
into defining an identity for this?
 
>- time-in-previous-state is oddly a string.  Why not a time of some sort?

<RR> I agree.

>- Future-proofing observation: An mpls-fec is not required to be an IP
>  address, although that is pretty much its only major manifestation at
>the
>  moment.  This may cause weirdness down the road.

<RR> So are you suggesting we use a choice too cater for this?

>- mpls-fec is of type ip-address.  FECs are typicall prefixes.

<RR> Ack.

>- bfd-diagnostic is missing code point 9. (See IANA issues.)

<RR> Ack.

>- Suggest adding the range restriction 0..31 to the description for
>  bfd-diagnostic.

<RR> Ack.

>- Naming elements such as nbor, conc, etc. are rather unnecessarily
>  truncated.  While I realize that yang is somewhat human readable, it's
>  mostly for machine interaction.  Reasonably expanded names make a module
>  easier to read, especially to English as a second language parties.
>This,
>  however, needs to be balanced against things like "admin" which is
>already
>  present in other BFD documents.

<RR> Ack.

>- In the bfd-auth-replay-protection description, please include citations
>to
>  the specific section of RFC 5880.

<RR> Ack.

>- The bfd-all-session grouping includes source-port and dest-port.  This
>may
>  not be relevant to all transports of BFD.  What discussion have you
>given
>  to restricting this, if any?

<RR> Good point. You¹re referring to encaps such as ACH I assume?

>- rx-ttl's description is a bit confusing to read.  I think you are trying
>  to suggest that the received TTL should be <= this value to be
>acceptable?

<RR> >= from what recall. We¹ll discuss this amongst authors.

>- potential nit on "ttl"; ipv6 calls this hop-count.  Any discussion on
>the
>  naming of this type among the authors?

<RR> None but we will.

>
>
>Features:
>Echo and demand mode are not pervasively implemented for BFD.  (I've not
>actually heard of an implementation of demand mode.)  I suspect these
>elements should be covered by if-feature?  While I realize the elements
>are
>optionally usable in the configuration leaves, the lack of a feature might
>lead a netconf implementation with the thought that they can use this
>feature.  What has been the author discussion on this point?

<RR> We have had some discussions on this. I agree that we should add
if-feature. Authors will discuss.

>
>IANA management of some module components:
>Several elements in this draft are things that will regularly need to be
>maintained.  Examples include:
>- bfd-path-type
>- bfd-auth-replay-protection
>- bfd-diagnostic
>
>I suspect we wish to place these components in a module/modules that are
>IANA maintained.  I'm a bit behind in current yang practices, but I
>suspect
>we have something like this defined now.  It unfortunately has the
>headache
>that it impacts name-spacing issues for the consuming modules.

<RR> Hadn¹t thought of that. I suspect BFD isn¹t the first module to need
thisŠ.

>
>Timing hints:
>Has any discussion been given to an operational module that might suggest
>what timer values are acceptable to an implementation?  Note: I'm not
>suggesting this as part of the core module, but it's a common "what timers
>does your implementation support" question that typically happens.

<RR> No discussions on this. We¹ll have to discuss. What is tricky with
this is that some implementations support different timer ranges based on
³path-type². E.g. 5 ms mayb be supported for single-hop but for multi-hop
the minimum could be 150 ms.

Regards,
Reshad.

>
>
>On Fri, Mar 10, 2017 at 09:29:45PM +0000, Reshad Rahman (rrahman) wrote:
>> Main changes from rev 04 are:
>> - Removed augment of network-instance, added reference to schema-mount
>> - bfd is not top-level anymore, augments control-plane-protocol from
>> RFC8022
>> - Addeds section on Interaction with other YANG modules
>> 
>> 
>> 
>> Regards,
>> BFD YANG authors.
>> 
>> On 2017-03-10, 4:26 PM, "Rtg-bfd on behalf of [email protected]"
>> <[email protected] on behalf of [email protected]> wrote:
>> 
>> >
>> >A New Internet-Draft is available from the on-line Internet-Drafts
>> >directories.
>> >This draft is a work item of the Bidirectional Forwarding Detection of
>> >the IETF.
>> >
>> >        Title           : Yang Data Model for Bidirectional Forwarding
>> >Detection (BFD)
>> >        Authors         : Reshad Rahman
>> >                          Lianshu Zheng
>> >                          Santosh Pallagatti
>> >                          Mahesh Jethanandani
>> >                          Greg Mirsky
>> >    Filename        : draft-ietf-bfd-yang-05.txt
>> >    Pages           : 53
>> >    Date            : 2017-03-10
>> >
>> >Abstract:
>> >   This document defines a YANG data model that can be used to
>>configure
>> >   and manage Bidirectional Forwarding Detection (BFD).
>> >
>> >
>> >
>> >The IETF datatracker status page for this draft is:
>> >https://datatracker.ietf.org/doc/draft-ietf-bfd-yang/
>> >
>> >There's also a htmlized version available at:
>> >https://tools.ietf.org/html/draft-ietf-bfd-yang-05
>> >
>> >A diff from the previous version is available at:
>> >https://www.ietf.org/rfcdiff?url2=draft-ietf-bfd-yang-05
>> >
>> >
>> >Please note that it may take a couple of minutes from the time of
>> >submission
>> >until the htmlized version and diff are available at tools.ietf.org.
>> >
>> >Internet-Drafts are also available by anonymous FTP at:
>> >ftp://ftp.ietf.org/internet-drafts/
>> >

Reply via email to