Re: [OPSAWG] Basic ietf process question ...

[Randy Presuhn]

Hi -

> From: "Andy Bierman" 
> To: "Romascanu, Dan (Dan)" 
> Cc: ; 
> Sent: Thursday, August 02, 2012 10:13 AM
> Subject: Re: [OPSAWG] Basic ietf process question ...
...
> NMS developers need to spend too many resources on translating
> naming and other data-modeling specific details so they can be
> usable within the application.  So if 1 data modeling language
> is not used, then deterministic, loss-less, round-trip translation
> between data modeling languages is needed.  Multiple
> protocols are not the problem -- incompatible data from multiple
> protocols is the problem.
...

Picking a single language or set of round-trip translatable languages
also isn't enough.  Its a fact of life that vendors will produce
models and implementations that are slightly, or even radically different.
The differences aren't necessarily even intentional, but nonetheless
introduce the need to talk about "similar" models and "operationally
equivalent" configurations, where the transformations needed to go
from what will do the job on one piece of equipment to what will
work on another may be substantial.  (From an implementation perspective
it might be better to think in terms of transformations necessary to go
from a common model of desired operational characteristics to the
dial tweaks and button pokes necessary to get a device to do the right
thing.)  Since great minds often think alike, even in the absence
of standards, there is not necessarily a formal "derived from" or
"subclass" or "common aspect" relationship between the definitions.
This may be an obvious use case for XSLT, but as far as I know nothing
has been done about *standardizing* such usage, other than discussions at the
IAB workshop oh-so-many years ago, and some ISO/ITU discussions in
the 1990s about eventual applications of the General Relationship
Model and the management domain/policy stuff in GDMO land.

Randy

Re: [OPSAWG] Basic ietf process question ...

[Juergen Schoenwaelder]

On Fri, Aug 03, 2012 at 04:31:42PM +0200, Robert Raszuk wrote:
> 
> But if we are at this phase I think creating a network elements
> abstraction layer and be able to configure/monitor any protocol and
> service at the unified way is one of the building blocks we should
> start with. And of course I think this is very clear to everyone if
> it is not made mandatory in each draft as new section or appendix it
> is just not going to happen in practice.
> 

Writing data models that are implementable on multiple platforms is
non trivial. We started working on your "building block" but it will
take time. And after all, a data model is only worth something if it
gets implemented.  A mandatory section does not help much with all of
this. What helps is people who commit time to work on it, who review
stuff, producing good and hopefully widely implementable models.

/js

-- 
Juergen Schoenwaelder   Jacobs University Bremen gGmbH
Phone: +49 421 200 3587 Campus Ring 1, 28759 Bremen, Germany
Fax:   +49 421 200 3103 

Re: [OPSAWG] Basic ietf process question ...

[Juergen Schoenwaelder]

On Fri, Aug 03, 2012 at 09:22:10AM +0200, Robert Raszuk wrote:
> 
> Aha .. so you are saying that MIBs are not mandatory  Very
> interesting. So I guess SSH to the routers and box by box cli
> provisioning is here to stay for a while I think :(
> 

Robert,

you may want to take a closer look at the data models currently being
defined in the NETMOD working group. Please review them and send any
comments.

/js

-- 
Juergen Schoenwaelder   Jacobs University Bremen gGmbH
Phone: +49 421 200 3587 Campus Ring 1, 28759 Bremen, Germany
Fax:   +49 421 200 3103 

Re: [OPSAWG] Basic ietf process question ...

[David Harrington]

+1

--
David Harrington
ietf...@comcast.net
+1-603-828-1401





On 8/2/12 12:59 PM, "Romascanu, Dan (Dan)"  wrote:

>Hi,
>
>The OPSAWG/OPSAREA open meeting this afternoon has an item on the agenda
>concerning the revision of RFC1052 and discussing a new architecture for
>management protocols.
>
>
>My personal take is that no one protocol, or one data modeling language
>can match the operational requirements to configure and manage the wide
>and wider range of hosts, routers and other network devices that are
>used to implement IP networks and protocols. We should be talking
>nowadays about a toolset rather than one tool that fits all. However,
>this is a discussion that just starts.
>
>Regards,
>
>Dan
>
>
>
>
>> -Original Message-
>> From: ietf-boun...@ietf.org [mailto:ietf-boun...@ietf.org] On Behalf
>Of
>> Robert Raszuk
>> Sent: Thursday, August 02, 2012 7:25 PM
>> Cc: ietf@ietf.org
>> Subject: Basic ietf process question ...
>> 
>> All,
>> 
>> IETF documents have number of mandatory sections .. IANA Actions,
>> Security Considerations, Refs, etc ...
>> 
>> Does anyone have a good reason why any new protocol definition or
>> enhancement does not have a build in mandatory "XML schema" section
>> which would allow to actually use such standards based enhancement in
>> vendor agnostic way ?
>> 
>> There is a lot of talk about reinventing APIs, building network wide
>OS
>> platform, delivering SDNs (whatever it means at any point of time for
>> one) ... but how about we start with something very basic yet IMHO
>> necessary to slowly begin thinking of network as one plane.
>> 
>> I understand that historically we had/still have SNMP however I have
>> never seen this being mandatory section of any standards track
>document.
>> Usually SNMP comes 5 years behind (if at all) making it obsolete by
>> design.
>> 
>> NETCONF is great and very flexible communication channel for
>> provisioning. However it is sufficient to just look at number of ops
>> lists to see that those who tried to use it quickly abandoned their
>> efforts due to complete lack of XML schema from each vendor they
>happen
>> to use or complete mismatch of vendor to vendor XML interpretation.
>> 
>> And while perhaps this is obvious I do not think that any new single
>> effort will address this. This has to be an atomic and integral part
>of
>> each WG's document.
>> 
>> Looking forward for insightful comments ...
>> 
>> Best,
>> R.
>> 
>
>___
>OPSAWG mailing list
>ops...@ietf.org
>https://www.ietf.org/mailman/listinfo/opsawg

Re: [OPSAWG] Basic ietf process question ...

[Andy Bierman]

On Thu, Aug 2, 2012 at 9:59 AM, Romascanu, Dan (Dan)  wrote:
> Hi,
>
> The OPSAWG/OPSAREA open meeting this afternoon has an item on the agenda
> concerning the revision of RFC1052 and discussing a new architecture for
> management protocols.
>
>
> My personal take is that no one protocol, or one data modeling language
> can match the operational requirements to configure and manage the wide
> and wider range of hosts, routers and other network devices that are
> used to implement IP networks and protocols. We should be talking
> nowadays about a toolset rather than one tool that fits all. However,
> this is a discussion that just starts.

NMS developers need to spend too many resources on translating
naming and other data-modeling specific details so they can be
usable within the application.  So if 1 data modeling language
is not used, then deterministic, loss-less, round-trip translation
between data modeling languages is needed.  Multiple
protocols are not the problem -- incompatible data from multiple
protocols is the problem.

>
> Regards,
>
> Dan
>

Andy

>
>
>
>> -Original Message-
>> From: ietf-boun...@ietf.org [mailto:ietf-boun...@ietf.org] On Behalf
> Of
>> Robert Raszuk
>> Sent: Thursday, August 02, 2012 7:25 PM
>> Cc: ietf@ietf.org
>> Subject: Basic ietf process question ...
>>
>> All,
>>
>> IETF documents have number of mandatory sections .. IANA Actions,
>> Security Considerations, Refs, etc ...
>>
>> Does anyone have a good reason why any new protocol definition or
>> enhancement does not have a build in mandatory "XML schema" section
>> which would allow to actually use such standards based enhancement in
>> vendor agnostic way ?
>>
>> There is a lot of talk about reinventing APIs, building network wide
> OS
>> platform, delivering SDNs (whatever it means at any point of time for
>> one) ... but how about we start with something very basic yet IMHO
>> necessary to slowly begin thinking of network as one plane.
>>
>> I understand that historically we had/still have SNMP however I have
>> never seen this being mandatory section of any standards track
> document.
>> Usually SNMP comes 5 years behind (if at all) making it obsolete by
>> design.
>>
>> NETCONF is great and very flexible communication channel for
>> provisioning. However it is sufficient to just look at number of ops
>> lists to see that those who tried to use it quickly abandoned their
>> efforts due to complete lack of XML schema from each vendor they
> happen
>> to use or complete mismatch of vendor to vendor XML interpretation.
>>
>> And while perhaps this is obvious I do not think that any new single
>> effort will address this. This has to be an atomic and integral part
> of
>> each WG's document.
>>
>> Looking forward for insightful comments ...
>>
>> Best,
>> R.
>>
>
> ___
> OPSAWG mailing list
> ops...@ietf.org
> https://www.ietf.org/mailman/listinfo/opsawg

Re: [OPSAWG] Basic ietf process question ...

[Robert Raszuk]

Hi Juergen,

Many thx for the great suggestion !

However perhaps you are much more knowledgeable in that area and could 
recommend which model fit the best the requirement to standardize 
configuration of any new protocol or protocol extension at least in the 
space of routing and routing protocols or services being based on them ?


As you know the current IRS framework driven by junisco is trying to 
come with common API to the routing system agreed across vendors. This 
is great as attempts never happened in the past.


But if we are at this phase I think creating a network elements 
abstraction layer and be able to configure/monitor any protocol and 
service at the unified way is one of the building blocks we should start 
with. And of course I think this is very clear to everyone if it is not 
made mandatory in each draft as new section or appendix it is just not 
going to happen in practice.


Best regards,
R.


On Fri, Aug 03, 2012 at 09:22:10AM +0200, Robert Raszuk wrote:


Aha .. so you are saying that MIBs are not mandatory  Very
interesting. So I guess SSH to the routers and box by box cli
provisioning is here to stay for a while I think :(



Robert,

you may want to take a closer look at the data models currently being
defined in the NETMOD working group. Please review them and send any
comments.

/js

RE: [OPSAWG] Basic ietf process question ...

[Romascanu, Dan (Dan)]

Yes. The question is whether a basic information model written in XML can be a 
useful starting point (trying to interpret the proposal made by Robert). 

Dan

> -Original Message-
> From: Andy Bierman [mailto:a...@yumaworks.com]
> Sent: Thursday, August 02, 2012 8:14 PM
> To: Romascanu, Dan (Dan)
> Cc: rob...@raszuk.net; ops...@ietf.org; ietf@ietf.org
> Subject: Re: [OPSAWG] Basic ietf process question ...
> 
> On Thu, Aug 2, 2012 at 9:59 AM, Romascanu, Dan (Dan)
>  wrote:
> > Hi,
> >
> > The OPSAWG/OPSAREA open meeting this afternoon has an item on the
> > agenda concerning the revision of RFC1052 and discussing a new
> > architecture for management protocols.
> >
> >
> > My personal take is that no one protocol, or one data modeling
> > language can match the operational requirements to configure and
> > manage the wide and wider range of hosts, routers and other network
> > devices that are used to implement IP networks and protocols. We
> > should be talking nowadays about a toolset rather than one tool that
> > fits all. However, this is a discussion that just starts.
> 
> NMS developers need to spend too many resources on translating naming
> and other data-modeling specific details so they can be usable within
> the application.  So if 1 data modeling language is not used, then
> deterministic, loss-less, round-trip translation between data modeling
> languages is needed.  Multiple protocols are not the problem --
> incompatible data from multiple protocols is the problem.
> 
> >
> > Regards,
> >
> > Dan
> >
> 
> Andy
> 
> >
> >
> >
> >> -Original Message-
> >> From: ietf-boun...@ietf.org [mailto:ietf-boun...@ietf.org] On Behalf
> > Of
> >> Robert Raszuk
> >> Sent: Thursday, August 02, 2012 7:25 PM
> >> Cc: ietf@ietf.org
> >> Subject: Basic ietf process question ...
> >>
> >> All,
> >>
> >> IETF documents have number of mandatory sections .. IANA Actions,
> >> Security Considerations, Refs, etc ...
> >>
> >> Does anyone have a good reason why any new protocol definition or
> >> enhancement does not have a build in mandatory "XML schema" section
> >> which would allow to actually use such standards based enhancement in
> >> vendor agnostic way ?
> >>
> >> There is a lot of talk about reinventing APIs, building network wide
> > OS
> >> platform, delivering SDNs (whatever it means at any point of time for
> >> one) ... but how about we start with something very basic yet IMHO
> >> necessary to slowly begin thinking of network as one plane.
> >>
> >> I understand that historically we had/still have SNMP however I have
> >> never seen this being mandatory section of any standards track
> > document.
> >> Usually SNMP comes 5 years behind (if at all) making it obsolete by
> >> design.
> >>
> >> NETCONF is great and very flexible communication channel for
> >> provisioning. However it is sufficient to just look at number of ops
> >> lists to see that those who tried to use it quickly abandoned their
> >> efforts due to complete lack of XML schema from each vendor they
> > happen
> >> to use or complete mismatch of vendor to vendor XML interpretation.
> >>
> >> And while perhaps this is obvious I do not think that any new single
> >> effort will address this. This has to be an atomic and integral part
> > of
> >> each WG's document.
> >>
> >> Looking forward for insightful comments ...
> >>
> >> Best,
> >> R.
> >>
> >
> > ___
> > OPSAWG mailing list
> > ops...@ietf.org
> > https://www.ietf.org/mailman/listinfo/opsawg