Re: [pcp] Comments on draft-ietf-pcp-server-selection-01

["Tirumaleswar Reddy (tireddy)" <[e-mail]>]

Hi Dave,

Please see inline

> -----Original Message-----
> From: Dave Thaler [mailto:[e-mail]]
> Sent: Thursday, August 08, 2013 7:58 PM
> To: Tirumaleswar Reddy (tireddy); Ted Lemon
> Cc: [e-mail]; [e-mail]; draft-ietf-pcp-server-
> [e-mail]
> Subject: RE: [pcp] Comments on draft-ietf-pcp-server-selection-01
> 
> 
> Tiru wrote:
> [...]
> > > If you buy Stuart's argument that communicating with just one IP
> > > address per
> >
> > If we consider the use-case http://tools.ietf.org/html/draft-ietf-pcp-
> server-
> > selection-01#appendix-A.1 there will be one PVD and two PCP server IP
> > addresses. If the ICE agent has ICE lite implementation, then it will have
> to
> > communicate with both the PCP controlled firewalls to permit incoming ICE
> > connectivity checks.
> 
> That's an interesting assertion but you didn't give any justification.
> 
> Stuart asserted that communicating with one is sufficient, and simpler for the
> client, because it's the PCP server's job to communicate with the other ones
> to permit incoming ICE connectivity checks.  If you disagree with that,
> please respond with a technical argument as to why having the client
> communicate with both is better than having one server communicate
> with the other.

Comparing the two approaches:

[1] PCP client will have to handle both the scenarios http://tools.ietf.org/html/draft-ietf-pcp-server-selection-01#appendix-A.1 (where both the Firewalls are managed by the same administrative domain) and http://tools.ietf.org/html/draft-patil-pcp-multihoming-00#section-5.1 (Network managed Firewall provided by each ISP). For the latter scenario PCP client would have to communicate with both the firewalls and there is no way for FW1 or FW2 to talk to each other. 

So if PCP client would already have the logic to handle multiple PCP-controlled Firewalls why make a special treatment just for http://tools.ietf.org/html/draft-ietf-pcp-server-selection-01#appendix-A.1 ?

[2] If it's compared with your suggestion then the PCP client will send all the PCP requests to only one of the routers (let's say rtr1 in appendix-A.1); rtr1 based on the source IP address of the PCP request will have to forward the request again to rtr2 causing additional delay, processing overhead.

-Tiru.

> 
> -Dave