[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: RFC 5006 status

On Fri, 19 Mar 2010 16:35:57 +0530
Syam Madanapalli <smadanapalli@gmail.com> wrote:

> What's the difference between running Stateless DHCPv6 vs carrying
> DHCP options in ND messages?
> 

A very small amount of RAM, CPU and a single packet. They way people
are persisting with this "we've got to turn RAs in to half-duplex
DHCPv6", you'd think those all those things were getting dramatically
more expensive.

I have a challenge for the RA for everything crowd. Firstly, however,
I'd like to quote Yogi Berra, or possibly Jan L. A. van de Snepscheut
or Albert Einstein (according to
http://en.wikiquote.org/wiki/Yogi_Berra),

"In theory there is no difference between theory and practice. In
practice there is."

I haven't seen anybody justify this RA for everything view
by saying, "I deployed RAs + stateless DHCPv6, and I had (one or more
of) (a) devices run out of RAM, (b) run out of CPU or (c) run out of
network bandwidth" or listed some significant problem they experienced
with the existing mechanisms. So as far as I can see, all the arguments are theoretical,
with no practice to support the argument.

The existing mechanisms work, both in theory, and in practice. I think
that means that the onus is on the RA for everything proponents to
*prove* where it doesn't work before anybody should spend valuable time
redesigning the existing mechanisms. There are better IPv6 things to
spend time on that are far more important to *get* working, rather than
spending time on changing the way existing things work.




> If nodes can implement DHCP options in ND messages, I could not
> understand why cannot they implement Stateless DHCPv6 itself.
> 
> In case ND option for DHCPv6 options is better than running stateless
> DHCPv6, then we need to look at whether there exists any other options
> other the DNS server address for which IETF is looking at defining RA
> options.
> 
> Thanks,
> Syam
> 
> On Fri, Mar 19, 2010 at 3:29 PM, Rémi Després <remi.despres@free.fr> wrote:
> >
> > Le 17 mars 2010 à 16:18, Fred Baker a écrit :
> >
> >> http://www.ietf.org/rfc/rfc5006.txt
> >>
> >> (1) Please take a look at the document in the next few days; if you have comments on it (eg, you think it should be changed in some way), please comment to v6ops.
> >
> > While supporting in general the idea, there is one improvement I suggest to make:
> > Rather than a specific RA option for Recursive DNS Servers, standardize the generic RA option to embed some stateless DHCP options, as proposed in draft-krishnan-intarea-ra-dhcp-00.
> >
> > This is in my understanding more powerful without being more complex.
> > By giving to routers a general possibility to broadcast stateless DHCP parameters (in addition to their still being obtainable in DHCPv6), not only the purpose of RFC50026 is achieved but, in one shot, the same progress is made for all common parameters that may concern all or most hosts.
> >
> > Hosts that support the RA DHCP option only have to: (1) process embedded DHCP options that they understand; (2) skip others; (3) request in DHCPv6 only options that aren't already received in RAs, if any.
> >
> > Note that Suresh Krishnan has 15min slot scheduled at the 6man meeting of Wednesday for:
> > "Stateless DHCPv6 and Router Advertisements for propagating configuration information".
> >
> > I add Suresh to the list, and Dan Wing who is known to support this approach.
> >
> > RD
> >
> >
>