Re: [v6ops] new draft: draft-taylor-v6ops-fragdrop

[Joel Jaeggli <[e-mail]>]

On Nov 1, 2012, at 12:56 PM, Joe Touch <[e-mail]>
 wrote:

> 
> 
> On 11/1/2012 12:24 PM, [e-mail] wrote:
>>> IMO, a router that looks beyond the IP headers is doing DPI. I realize
>>> others think DPI means looking at the E2E data.
>>> 
>>> However, to avoid confusion, I'll just state that:
>>> 
>>> - anything that looks only at an IP packet is a router
>>> 
>>> - anything that looks at higher layer headers isn't a router anymore. it
>>> might be a firewall, filter, etc., and whether the box can do that at
>>> line rate is a question customers should ask their vendors, not a
>>> problem the IETF should solve.
>> 
>> This looks like a nice theoretical classification that is rather useless
>> in practice.
>> 
>> A modern router with multiple 10G interfaces (say a Juniper MX or Cisco
>> ASR9K) can be configured to look at higher layers. The primary function
>> of these routers is forwarding packets at high speeds. The fact that
>> these routers *can* look at higher layers is not going to make me stop
>> calling them routers. I don't think I'm alone here :-)
> 
> Fine.

not really it isn't. What got me down this rathole was l3+l4 hashing.

> Then either get your vendors - of whatever boxes you want to call them - to support the current option structure, or choose not to support IPv6.

That's not an operationally useful statement. and it's not what's cooked into silicon by a number of companies that are well represented at the IETF.

> There are choices here.

Indeed, and the observable phenomena on the internet indicate what operators and their vendors arrived at indepedantly. 

> Joe