Re: [Idr] Unknown Attributes seen in the wild

Just to clarify, it's not just about sending an attribute, its also about receiving it.

Thanks,
Jakob.

On Oct 30, 2016, at 2:08 PM, Jakob Heitz (jheitz) <jheitz@cisco.com<mailto:jheitz@cisco.com>> wrote:

Robert,

That does not answer my question.

Thanks,
Jakob.

On Oct 30, 2016, at 1:37 PM, Robert Raszuk <robert@raszuk.net<mailto:robert@raszuk.net>> wrote:

Those self invented attributes will not be passed/leaked as such safi they are to be part of will not be capability negotiated in the first place with any Internet router.

Thx,
R.

On Sun, Oct 30, 2016 at 9:26 PM, Jakob Heitz (jheitz) <jheitz@cisco.com<mailto:jheitz@cisco.com>> wrote:
Send them one of those attributes in a different safi. What do you think will happen?
Will they pass it on harmlessly, like any unknown optional transitive attribute?

Thanks,
Jakob.

On Oct 30, 2016, at 1:05 PM, Robert Raszuk <robert@raszuk.net<mailto:robert@raszuk.net>> wrote:

Jakob,

contained != restricted

Moreover there is number of overlay solutions in the wild which run under the cover BGP with various self defined attributes and safis. Just think about SD-WAN or DC overlay solutions.

Some of them tried to take IETF path some never bothered seeing how hard is to get early IANA allocations or being aware of multiple implementation requirement.

Some are limited to contained deployments some are encrypted. It is close to impossible to now list all of those.

It was pretty clear this is coming ... one strategic solution was an attempt 6 years back to separate BGP into BGP used for transport for all of the application and services (new BGP port) and BGP used for Internet routing (179). That would allow complete separation of the two.

https://tools.ietf.org/html/draft-raszuk-ti-bgp-01

Thx,
r.

On Sun, Oct 30, 2016 at 8:40 PM, Jakob Heitz (jheitz) <jheitz@cisco.com<mailto:jheitz@cisco.com>> wrote:
There is no concept of a BGP attribute being "restricted" to a safi. If that were the case, then you could use a single attribute code for completely different attributes in different safis. An attribute that is understood in one safi, but nonsense in a second safi will cause a malformed update rather than an unknown attribute when received in the second safi.

Thanks,
Jakob.

> On Oct 30, 2016, at 12:16 PM, Colin Petrie <colin@spakka.net<mailto:colin@spakka.net>> wrote:
>
>> On 30/10/16 19:05, Robert Raszuk wrote:
>> The good news is that they should not be leaked as they are to be send
>> in their dedicated SAFIs so unless peer exchanges right capabilities
>> they should be quite contained.
>
> In this case, we see them appear on BGP sessions that only negotiate
> IPv4/Unicast or IPv6/Unicast.
>
> Cheers,
> Colin
>
> _______________________________________________
> Idr mailing list
> Idr@ietf.org<mailto:Idr@ietf.org>
> https://www.ietf.org/mailman/listinfo/idr

_______________________________________________
Idr mailing list
Idr@ietf.org<mailto:Idr@ietf.org>
https://www.ietf.org/mailman/listinfo/idr