Re: [Isis-wg] draft-amante-isis-reverse-metric-01

[Shane Amante <[e-mail]>]

Robert, Bruno, (Jay),

It is an explicit /non-goal/, even outside the context of this draft, to port an NMS system and/or configuration generation system *into* IS-IS.  Call me old-school, but a routing protocol, (particularly an IGP), should only distribute topology/reachability information, as fast as [is, safely] possible.  Asking that routing protocol to also push around counters (snmp over IS-IS <barf>), general-purpose configuration information (netconf over IS-IS <barf>), etc. is contrary to that goal for all the reasons that you're already familiar with.  :-)

To be more clear, the scope of this draft is to make operation of an IS-IS network dramatically less error-prone by simplifying the handling of _a_ (notice the explicit use of the singular) very well-known, widely-used, routine maintenance procedure that occurs on networks day-in and day-out, the world over, _specifically_: p2p link /and/ multi-access LAN isolation.

Thanks,

-shane


On Dec 7, 2010, at 15:46 MST, Robert Raszuk wrote:
> Hi Jay,
> 
> Just one clarification ...
> 
> What I and also Bruno pointed out was a bit more restrictive that your interpretation :)
> 
> While each node would be an equal class citizen in such ability to configure the network or to set a parameter of a p2p or p2mp link it does not mean that one would essentially need to enable such ability on each node.
> 
> Clearly also it would not be available to anyone. Only to the same access level as today allowing for configuration.
> 
> As to the topic of what protocol would fit such role I do agree with your comment that we would probably need some form of hybrid. The reliable domain wide flooding should be combined with incremental updates ability.
> 
> Volume wise I am not sure we are to have "huge volumes". Clearly what differs this type of fundamentally different approach to NMS that it is not targetted to be a carrier for SNMP pooling stats or netflow records.
> 
> Many thx,
> R.
> 
>> wide config making each router to be an NMS station. Call it C-ISIS.
>> Example: instead of configuring links on the adjacent nodes configure
>> link parameter anywhere, instead of configuring node at a node ..
>> configure it anywhere.
>> 
>> Very interesting. Configure anywhere by anyone (maybe any computer)
>> will be every powerful and very scaring too (security). A graph of
>> network and transport of configure information will be needed. If
>> protocol is proposed to do the job, should be a single protocol best
>> fit? ISIS does know the topology of the network, on another hand BGP
>> is better to transmit huge data and does incremental updates
>> reliablely. A combination of functions from multiple protocol seems
>> better than extending a single protocol.
>> 
>> Jay
>> 
>> 
>> Many thx, R.
>> 
>> PS.
>> 
>> /* After all we have many uses for ISIS today ... TRILL included :)
>> */
>> 
>> 
>>> Robert,
>>> 
>>>> Instead of working on network wide level config tool/protocol we
>>>> are extending protocols to achieve the equivalent functionality.
>>>> And this is not only in ISIS, we do the same in BGP, we do the
>>>> same in MPLS etc ...
>>> 
>>> 
>>> So, you're suggesting that we stop work on improving protocols
>>> until the NMS folks get their act together.
>>> 
>>> This doesn't seem like a practical suggestion.
>>> 
>>> Regards, Tony
>>> 
>>> 
>> 
>> _______________________________________________ Isis-wg mailing list
>> [e-mail] https://www.ietf.org/mailman/listinfo/isis-wg
>> 
> 
> _______________________________________________
> Isis-wg mailing list
> [e-mail]
> https://www.ietf.org/mailman/listinfo/isis-wg