Re: [i2rs] I-D Action: draft-keyupate-i2rs-bgp-usecases-02.txt

[t.petch <[e-mail]>]

----- Original Message -----
From: "Dean Bogdanovic" <[e-mail]>
To: "Susan Hares" <[e-mail]>
Sent: Friday, June 13, 2014 11:06 PM

Susan,

My answer  to your question
do we ever let I2RS upon a command transfer policies to the persistent
storage?

My initial answer is no. And reason is security. Network admins want to
know exact state of the device after rebooting. If we want to allow
transfer of policies, then we would have to define roles and which roles
would be allowed to do that transfer. We are making things more complex,
when there are existing mechanisms for admins to do that.

<tp>
The counter argument to this is having got it right once, for some,
perhaps large, instantiation of an I2RS it, do you want to have to do it
all over again?

That is, I agree that after reboot, an operator should know that the box
is in the state defined by the NETCONF startup configuration datastore
with nothing added but that there may be a valid case for saying that
all the changes downloaded by I2RS remain stored, but not applied, eg in
a candidate I2RS datastore, which can then be copied to the running I2RS
datastore when the operator is satisfied that the time is right.

Just a thought.

Tom Petch









Dean

On Jun 13, 2014, at 5:49 PM, Susan Hares
<[e-mail]<mailto:[e-mail]>> wrote:

Dean:
Thank you for your thoughtful answer.  I was looking for it, but I’m
glad you watched Croatia (I’m sorry about Croatia’s loss).

I agree with the types of policies and the status: persistent,
transient, and ephemeral.  However, do we ever let I2RS upon a command
transfer policies to the persistent storage?  This what I read in REQ04.
In reading the email, I’m not sure how to summarize the WG’s approach.
My answer would be “no”, but if this is a WG document the answer needs
to come from the WG.

Sue

From: i2rs [mailto:[e-mail]<mailto:[e-mail]>] On
Behalf Of Dean Bogdanovic
Sent: Friday, June 13, 2014 2:44 PM
To: Susan Hares
Cc: <[e-mail]<mailto:[e-mail]>>; Susan Hares;
<[e-mail]<mailto:[e-mail]>>; t.petch; Keyur Patel (keyupate);
Hannes Gredler; Russ White
Subject: Re: [i2rs] I-D Action: draft-keyupate-i2rs-bgp-usecases-02.txt

Susan,

Sorry for late reply, but yesterday started a very significant
quadrennial event (FIFA World Cup) and Croatia played (and lost with
help of the referee).

WRT REQ04, I agree with the posts and here are few thoughts on this

We should try to divide policies into few categories.

1. persistent
There is a set of policies that have to be available from very start on
the device. Those policies should be persistent on the device and I see
them changing infrequently. IMO, there is no need for I2RS to manage
those policies, readonly access is sufficient.
2. transient
Policies that are temporary defining some fwd behavior of device. I can
see lot of cases where different applications based on some network
conditions want to change forwarding behavior. Those should not be
available after reboot.
3. locally defined
by this I mean policies that defined by admin, applications through
local I2RS agent. These can be transient and persistent, where I would
classify that I2RS agent policies are only transient. Actually after
rereading this, I would even consider policies defined by I2RS as
remotely defined and therefore transient.
4. remotely defined
by this I mean policies pushed from a different device (policy server,
router) via some protocol (DIAMETER, RADIUS, BGP). IMO, those should be
always ephemeral.

Dean

On Jun 11, 2014, at 9:08 PM, Susan Hares
<[e-mail]<mailto:[e-mail]>> wrote:


Dean:

I combined REQ01/02 and REQ08/09.  I've put the requirements in the
front of the text.  Please let know if have any suggestions on these
approved changes.  I wait 24 hours, and then spin the draft.

On the agreement on REQ04, I cannot find a firm consensus.  I would ask
Jeff Haas or Ed Crabbe to indicate if they think there is a consensus on
the WG. I highlight a few messages below. The document is proposed for
WG consensus so I will change it if the WG has consensus.


Sue Hares


Search for Consensus
=====
Based on your comment, I sent looking for WG Direction regarding BGP or
I2RS putting state.   I cannot find it.  BGP has a Flow specification
(RFC5575).  Where do you think those flow specifications end up?
Writing into runtime configuration state? Writing into something like
I2RS running data store?  BGP ORFs might be kept in the BGP state or in
associated features (Add/delete) in BGP, but Flow specifications are
targeted toward data flow.

On the list I could find the following:

1. I2RS BGP state to configuration - Wes George (operator) makes a
comment that I2RS configuration should not replace current configuration
related to BGP.

http://www.ietf.org/mail-archive/web/i2rs/current/msg00826.html



2. There is the Architecture Discussion 2: Persistence (ephemeral vs.
permanent) - is the debate for the architecture document regarding
keeping state in the I2RS
Begin:
http://www.ietf.org/mail-archive/web/i2rs/current/msg01027.html

Joel's: no state across a reboot:
http://www.ietf.org/mail-archive/web/i2rs/current/msg01034.html


3. Wes George (operator) makes a comment that I2RS configuration should
not replace current configuration related to BGP.
http://www.ietf.org/mail-archive/web/i2rs/current/msg00826.html


There is the Architecture Discussion 2: Persistence (ephemeral vs.
permanent)
http://www.ietf.org/mail-archive/web/i2rs/current/msg01027.html


Multiple clients writing to agents (raised by Himanshu Shah)
http://www.ietf.org/mail-archive/web/i2rs/current/msg01139.html


Jeff (chair hat off) states he does not want to have I2rs changing state
tables come from routing protocols (BGP--> I2RS state).  He also feel
dynamic state tables should be read-only, and not writable as suggested
by the use case.
http://www.ietf.org/mail-archive/web/i2rs/current/msg01666.html


In the same thread, Sri states the I2RS agent should not provide an
interface to change a table if there is no use-case to support it.
Dynamic protocols --> I2RS (I2RS read only).  I2RS--> RIB-IM.   Sri
states " I am yet to see a use-case that requires direct manipulation of
a single dynamic routing-protocol-instance specific route table by
something other than that protocol. I don't believe there should be any
such case."    However, here it has been in the BGP use case.
http://www.ietf.org/mail-archive/web/i2rs/current/msg01671.html


Jeff responds to Sri in tends to agree and does not mention the use
case.
http://www.ietf.org/mail-archive/web/i2rs/current/msg01752.html


Sue Hares

-----Original Message-----
From: Dean Bogdanovic [mailto:[e-mail]<http://juniper.net>]
Sent: Friday, June 06, 2014 4:03 PM
To: Susan Hares
Cc: t.petch; <[e-mail]<mailto:[e-mail]>>; Keyur Patel
(keyupate); Hannes Gredler; Russ White; Susan Hares;
<[e-mail]<mailto:[e-mail]>>
Subject: Re: [i2rs] I-D Action: draft-keyupate-i2rs-bgp-usecases-02.txt

Susan,

Many people don't know what NLRI abbreviation stands for (Network Layer
Reachability Information , so writing it out first time would be a good
idea.

Throughout the text, the requirement number sequence is confusing until
you get to the very and where all requirements are listed and then it
makes sense.

REQ04: The ability to interact with various policies configured on
      the forwarding devices, in order to inform the policies
      implemented by the dynamic routing processes.  This interaction
      should be through existing configuration mechanisms, such as
      NETCONF, and should be recorded in the configuration of the local
      device so operators are aware of the full policy implemented in
      the network from the running configuration.
It is not clear to me if your requirement is that dynamic protocols
should impose persistent policies? It says it should be recorded in the
configuration of the local device.

I agree that those policies should be visible to operators and other
applications, but not sure if dynamic protocols should be allowed to
implement persistent policies. IMO, those should be ephemeral policies.
So maybe text should look like this
This interaction should be through existing configuration mechanisms,
such as NETCONF, and should be recorded in the running or ephemeral
configuration of the local device so operators are aware of the full
policy implemented in the network from the running configuration.

I'm trying to see major difference between REQ01/REQ02 and REQ08/REQ09?

In general I'm not sure if changing entries by dynamic protocol in RIB
is a good idea. If you plan to change only what is configured on the
local device, then that is OK, but if you start changing entries that
are pushed from other devices in the network, the system would get
unstable. And it looks to me that REQ09 would allow that.

Dean


On Jun 5, 2014, at 4:47 AM, Susan Hares
<[e-mail]<mailto:[e-mail]>> wrote:

> Tom:
>
> I'm glad to change the citation in the abstract.    On the authors,
this was
> merge of two drafts.
>
> Sue
>
> -----Original Message-----
> From: t.petch [mailto:[e-mail]<http://btconnect.com>]
> Sent: Thursday, June 05, 2014 4:35 AM
> To: Susan Hares; [e-mail]<mailto:[e-mail]>
> Cc: 'Keyur Patel (keyupate)'; Hannes Gredler; Russ White; 'Susan
> Hares'; [e-mail]<mailto:[e-mail]>
> Subject: Re: [i2rs] FW: I-D Action:
> draft-keyupate-i2rs-bgp-usecases-02.txt
>
> Sue
>
> Currently you have six authors which is too many for an RFC -
someone's
> got to go!   For me, this is not just an admin point - when
commenting,
> I like to have one or two names, no more, as the clear pen holders
> whom I can expect to act.  Too often, with so many names, everyone
> thinks that someone else will do something and nothing happens, so, in
> all seriousness, I oppose adoption until you sort this out amongst
yourselves.
>
> Note too that you have a citation in the Abstract, again not allowed -
> this can be surprising difficult to get round but get round it you,
> one or more thereof, must.
>
> Tom Petch
>
>
> ----- Original Message -----
> From: "Susan Hares" <[e-mail]<mailto:[e-mail]>>
> To: <[e-mail]<mailto:[e-mail]>>
> Cc: "'Keyur Patel (keyupate)'"
<[e-mail]<mailto:[e-mail]>>; "Hannes Gredler"
> <[e-mail]<mailto:[e-mail]>>; "Russ White"
<[e-mail]<mailto:[e-mail]>>; "'Susan Hares'"
> <[e-mail]<mailto:[e-mail]>>;
<[e-mail]<mailto:[e-mail]>>
> Sent: Wednesday, June 04, 2014 7:49 PM
> Subject: [i2rs] FW: I-D Action:
> draft-keyupate-i2rs-bgp-usecases-02.txt
>
>
>> Jeff and Ed:
>>
>> This updated draft has all the changes that Keyur Patel promised and
> updates
>> to the reference the current i2rs internet drafts.
>>
>> Would you please do a Working Group adoption call?
>>
>> Thank you,
>> Sue Hares
>>
>>
>> -----Original Message-----
>> From: i2rs [mailto:[e-mail]<mailto:[e-mail]>] On
Behalf Of
>> [e-mail]<mailto:[e-mail]>
>> Sent: Wednesday, June 04, 2014 1:44 PM
>> To: [e-mail]<mailto:[e-mail]>
>> Cc: [e-mail]<mailto:[e-mail]>
>> Subject: [i2rs] I-D Action: draft-keyupate-i2rs-bgp-usecases-02.txt
>>
>>
>> A New Internet-Draft is available from the on-line Internet-Drafts
>> directories.
>> This draft is a work item of the Interface to the Routing System
> Working
>> Group of the IETF.
>>
>>        Title           : Use Cases for an Interface to BGP Protocol
>>        Authors         : Keyur Patel
>>                          Rex Fernando
>>                          Hannes Gredler
>>                          Shane Amante
>>                          Russ White
>>                          Susan Hares
>> Filename        : draft-keyupate-i2rs-bgp-usecases-02.txt
>> Pages           : 17
>> Date            : 2014-06-04
>>
>> Abstract:
>>   A network routing protocol like BGP is typically configured and
>>   analyzed through some form of Command Line Interface (CLI) or
>>   NETCONF.  These interactions to control BGP and diagnose its
>>   operation encompass: configuration of protocol parameters, display
> of
>>   protocol data, setting of certain protocol state and debugging of
> the
>>   protocol.
>>
>>   Interface to the Routing System's (I2RS) Programmatic interfaces,
> as
>>   defined in draft-ietf-i2rs-architecture, provides an alternate way
> to
>>   control and diagnose the operation of the BGP protocol.  I2RS may
> be
>>   used for the configuration, manipulation, analyzing or collecting
> the
>>   protocol data.  This document describes set of use cases for which
>>   I2RS can be used for BGP protocol.  It is intended to provide a
> base
>>   for the solution draft describing a set of interfaces to the BGP
>>   protocol.
>>
>>
>> The IETF datatracker status page for this draft is:
>> https://datatracker.ietf.org/doc/draft-keyupate-i2rs-bgp-usecases/
>>
>> There's also a htmlized version available at:
>> http://tools.ietf.org/html/draft-keyupate-i2rs-bgp-usecases-02
>>
>> A diff from the previous version is available at:
>> http://www.ietf.org/rfcdiff?url2=draft-keyupate-i2rs-bgp-usecases-02
>>
>>
>> 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<http://tools.ietf.org>.
>>
>> Internet-Drafts are also available by anonymous FTP at:
>> ftp://ftp.ietf.org/internet-drafts/
>>
>> _______________________________________________
>> i2rs mailing list
>> [e-mail]<mailto:[e-mail]>
>> https://www.ietf.org/mailman/listinfo/i2rs
>>
>> _______________________________________________
>> i2rs mailing list
>> [e-mail]<mailto:[e-mail]>
>> https://www.ietf.org/mailman/listinfo/i2rs
>>
>
>
> _______________________________________________
> i2rs mailing list
> [e-mail]<mailto:[e-mail]>
> https://www.ietf.org/mailman/listinfo/i2rs

<draft-keyupate-i2rs-bgp-usecases-03.txt><draft-keyupate-i2rs-bgp-usecas
es-03.txt.pdf>