Re: [Idr] BGP autoconfiguration - draft-minto-idr-bgp-autodiscovery

Robert: 

The L3DN only requires IPv4 or IPv6 native.  BGP, as a protocol only uses IPv4 or IPv6 native. 

I am comforted to hear that my use case is clearer.   I realize that most MPLS interfaces support IP routing/switching (for about 20 years).  Please note that it is not the protocol BGP that is enforcing the policy,  but the implementation selecting particular interface.    The encapsulation information is optional.   

If you wish to remove this option, it is a useful input to the authors.  

Thanks for your feedback. 

Sue 

From: Robert Raszuk [mailto:robert@raszuk.net] 
Sent: Sunday, March 20, 2022 7:21 PM
To: Susan Hares
Cc: Jeffrey Haas; idr@ietf. org; draft-minto-idr-bgp-autodiscovery@ietf.org; Minto Jeyananth
Subject: Re: [Idr] BGP autoconfiguration - draft-minto-idr-bgp-autodiscovery

Sue,

What encapsulation happens to be used by BGP or for that matter any other protocol is (and should be) protocol agnostic. 

I hope we are not trying here to bind a TCP session to encapsulation X or Y irrespective of what routing (and given encap orchestration - LDP FEC binding or SR policy or something else) tells you to do. That would quickly get very very inconsistent and prone to many failures. 

Btw there is no such a thing as "MPLS interface". All MPLS interfaces I have seen do support IP routing/switching too. And honestly it may be much better if BGP keeps its transport limited to IPv4 and IPv6 native. 

Thx,

Robert

On Mon, Mar 21, 2022 at 12:02 AM Susan Hares <shares@ndzh.com> wrote:

Jeff: 

I'm really trying to answer your question - so if I do not understand your
question.  I'll try again. 

The BGP transport sessions can have policy that requires the BGP hello to be
exchanged across a L3 connection running on a particular MPLS interface.
This policy ties down the TCP to a specific L3 connection on a specific L2.5
connection.   This example, show the benefit for some overlay/underlay
deployments.  

How does it impact the BGP FSM?   You will not get an TCP connection up due
to machine policy (it is reject except for the l3 connection attached to
specific MPLS interface).     

I really hope I've understood your question. 

Sue 

-----Original Message-----
From: Jeffrey Haas [mailto:jhaas@pfrc.org] 
Sent: Sunday, March 20, 2022 2:39 PM
To: Susan Hares
Cc: 'Minto Jeyananth'; idr@ietf.org;
draft-minto-idr-bgp-autodiscovery@ietf.org
Subject: Re: [Idr] BGP autoconfiguration - draft-minto-idr-bgp-autodiscovery

Sue,

I'm asking a very specific question.  Let me repeat:

On Sun, Mar 20, 2022 at 02:33:56PM -0400, Susan Hares wrote:
> You asked  L3DN is useful where besides the DC.

No, I asked:
> > It is still unclear what the "encapsulation" fields are used for with
> > respect to BGP discovery.

> I am looking for the same level of answer from the draft-minto-*.   Are
they
> planning to use this for WAN BGP auto-configuration.  

I'm not speaking for them.  Minto or Shiva can commment on their draft.

> On the encapsulation,  as Randy has said AFI/SAFI.  Consider a data center
> with multiple overlays (L2, mpls, L3) and AFI/SAFIs.  Suppose, you want to
> use an MPLS interface for an L3 connections.  You might want to select a
> specific set of labels for the underlay for L3DN. 

That's still not clear.  

How is this used to setup BGP's transport session during discovery?
I.e. how does this impact BGP's TCP SYN and the rest of the session?

-- Jeff

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