Re: [pim] Replacing unicast BSMs with multicast

[Stig Venaas <stig.venaas at uninett.no>]

Cao Wei wrote:
> Hi, John
>    I can not give the accurate number that how many PIM routers are running in the network. 
> But as I know,  BSR/BSMs is more popular static RP. A colleague has years experience
> on network deployment  agree with me.

Right, so question is really not about BSR/BSMs but unicast BSMs.

>    
>    I think most of the vendors have implement Unicast BSM because previous version PIM-SM 
> (RFC 2362, I mean) require this and it has been there for years. So I doubt unicast BSM is 
> really a complex matter.  Another question we have to consider: is this  a so important 
> problem worth the risk of compromising the compatibility? And it should be noted that, up to
> now, there is not a perfact method can handle all cases without any cons. So I prefer keep
> unicast BSMs.

I've been thinking a bit more about compatibility. A router starting up, 
would/should get unicast BSM (or the equivalent) on each interface (or 
at least each link) that it's connected to where there are neighbours 
with BSR information. If some routers do unicast BSMs, and some the 
possible multicast replacement, then if one of the neighbours is a DR 
and implement same as you, then you're ok. In worst case you'll have to 
wait for the periodic BSR. It might also be possible to reduce the BSR 
period...

If we were to specify a replacement for unicast BSMs, then it could 
still be possible for vendors to have backwards compatibility and send 
unicast BSMs as well (perhaps a user configurable option), and also 
accept unicast BSMs, if they so wish. That would add complexity, but 
hopefully something that can disappear over time.

Ignoring the backwards compatibility, I believe using multicast is a far 
better solution, and I would like to make a good choice for the future 
(something we may need to live with for many many years to come). 
Regarding backwards compatibility, it's useful to see how many have 
implemented it, so we know how big an issue it is, and also consider how 
bad or well it still can work if you mix implementations (as I tried to 
do in the first paragraph above).

It's true that unicast BSMs have been in 2362 for a long time, but there 
are still implementations that don't support it.

It would be nice to get input from more people that have implemented it. 
 Also, IMO, who has not implemented it, and if possible why. In 
particular, would a multicast replacement be more likely to be implemented?

Stig

> Regards,
> Cao Wei 
>
> ----- Original Message ----- 
> From: "John Zwiebel" <jzwiebel at cisco.com>
> To: "Dino Farinacci" <dino at cisco.com>
> Cc: <mmcbride at cisco.com>; <pim at ietf.org>
> Sent: Wednesday, April 26, 2006 6:17 AM
> Subject: Re: [pim] Replacing unicast BSMs with multicast
>
>
>
> > On Apr 25, 2006, at 3:04 PM, Dino Farinacci wrote:
> >
> >
> > >    Could we do a survey on how widespread BSR/BSMs are used to  
> > > quantify the
> > >    impact on compatibility?
> >
> > That's unicast BSMs and how many routers in the real world actually
> > have code for which it works.  Not whether or not a specific  
> > implementation
> > has deployed it.
> >
> > Although another question to ask might be, when was it first deployed
> > by a given vendor.
> >
> > Thanks
> >
> > _______________________________________________
> > pim mailing list
> > pim at ietf.org
> > https://www1.ietf.org/mailman/listinfo/pim


_______________________________________________
pim mailing list
pim at ietf.org
https://www1.ietf.org/mailman/listinfo/pim