[c-nsp] flow spec

[Yann Berthier]

On Sun, 03 Dec 2006, at 17:06, Richard A Steenbergen wrote:

> On Sun, Dec 03, 2006 at 12:06:22PM -0700, Sergio D. wrote:
> > Has Cisco implemented this yet?
> > http://www.tcb.net/draft-marques-idr-flow-spec-00.txt
> 
> My understanding is that for some unknown reason Cisco flatly refuses to 
> add this to BGP, and wants to implement a new protocol.

   btw, what i've heard recently from another cisco guy was perfectly
   aligned with what Roland Dobbins just stated in his reply - perhaps
   there has been some shift in cisco's position recently regarding flow
   spec ?

> The only thing I've heard which might hint to a reason is a belief that 
> "complicating BGP with more features" is bad, which seems naive and 
> outdated. In a modern multiprotocol BGP world all we're left with is a 
> mature and well tested protocol for distributing NLRIs and attributes, 
> cleanly seperated by AFI/SAFI. There is no real reason not use to BGP to 
> transmit filtering information as flowspec does, but a lot of reasons why 
> you should.
> 
> Here's to hoping Cisco catches up to reality on this one. I used the heck 
> out of flow-spec on Juniper's and it is a great way to handle distributed 
> filtering. My only regret is that I couldn't get more customers to use it 
> to feed me filter information for their own blocks. Maybe if it had 
> support in zebra/quagga/whatever kids with linux boxes are using these 
> days... :)

   Indeed. In the whatever category, Henning Brauer (OpenBGP author)
   statement was that he had no direct interest in supporting it, but
   some interested party may still propose to fund an implementation for
   OpenBGP ... 

      - yann