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

[Lorenzo Colitti <[e-mail]>]

On Wed, Oct 31, 2012 at 11:47 PM, Nick Hilliard <[e-mail]> wrote:

> On 31/10/2012 14:25, Randy Bush wrote:
> > and how much of good people's time do you plan to waste on this
> > windmill, don?
>
> Randy,
>
> you're right that we appear to have backed ourselves into a corner where we
> expect kit to be able to handle TLV processing at wire speed.
>
> Rather than stowing thrones, do you have any constructive suggestions on
> how to deal with the situation?
>

How about deprecate fragmentation at the IP layer and standardize a
higher-layer alternative?

IPv6 extension headers were standardized in 1995. I'd argue that in 1995
nobody could have foreseen hardware forwarding and that we'd have routers
capable of pushing tens of terabits per second. And while other aspects of
IPv6 were designed to reduce load on routers,. unfortunately, extension
headers were not. Because routers *do* inspect packet contents - for
security, QoS classification, accounting, and a thousand other
reasons. From the point of view of reducing load on routers, extension
headers were a mistake.

Yes, it's possible to put arbitrary length TLV processing in silicon so it
can forward at line rate packets with 8 extension headers - but we're not
suggesting that router vendors do that, are we? Because silicon is
expensive. And if that's not what we are suggesting, then what is it that
we really want? Up to one extension header? Up to two?