Re: [Autoconf] DHCP and AUTOCONF (was: Call for comments to a new AUTOCONF charter proposal.)

[Alexandru Petrescu <[e-mail]>]

Hi Emmanuel,

Le 03/07/2010 12:17, Emmanuel Baccelli a écrit :
> Hi Charlie,
>
> I agree with you about trying to avoid the dichotomy
> centralized/decentralized.

I agree to try to avoid the dichotomy centralized-decentralized because
DHCP has already many features exhibiting decentralization:

-DHCPv6 dst address is multicast: several servers can be in that group.
-several DHCP Servers running on the the same address with VRRP.
-Several Relays per a DHCP Server.
-DHCP Reconfigure message to let a Server suggest a Client another
  Server.
-RA tell the Node there's a better DHCP Server available.
-there are surely more.

> If some people want to work on a DHCP-dependent solution, let them
> try. If some other people want to work in parallel on a
> DHCP-independent solution, why not let them do it too?

I don't know, it's too simply put.  I'd evaluate the work quantity
needed for each, and prefer reuse.

Alex

>
> In my mind, these two solutions could be pretty complementary.
>
> Emmanuel
>
>
>
> On Fri, Jul 2, 2010 at 6:41 PM, Charles E. Perkins
> <[e-mail]
> <mailto:[e-mail]>> wrote:
>
> Hello Emmanuel and all,
>
> Thanks for supporting the consideration of decentralized approach.
> I'm pretty sure we could get at least one good working solution in
> time a lot faster than it took to agree that not everything is an
> Ethernet.
>
> However, it's important to avoid making a sharp dichotomy between
> "centralized" and "decentralized" approaches.  My big concern was
> that somehow the DHCP model was going to be considered the only
> viable choice because people (outsiders?) consider it to be the only
>  known quantity.  There are other ways to have more centralized or
> less centralized procedures, with or without proxy assistance,
> perhaps hybridized approaches, and with improved availability via
> elections.
>
> Those are just a few of the options, and probably all of them are way
> better than trying to shoehorn DHCP where it does not seem to fit.
> Flushing years of development and wisdom down the toilet just because
> they doesn't spell DHCP seems really wrong to me.
>
> Another point to keep in mind -- DHCP was designed and built from day
> one to be a _managed_ solution for autoconfiguration. It seems quite
> clear to me that most of the interesting development and inspiration
> for ad hoc networks has gone towards enabling networking in
> environments where no such tightly managed administration is
> possible. Thus, in my view, DHCP almost tautologically disqualifies
> itself from consideration without major structural redesign.  How is
> it, then, that DHCP keeps cropping up in the discussion?
>
> Regards, Charlie P.
>
>
>
>
> On 7/1/2010 2:28 AM, Emmanuel Baccelli wrote:
>
>
>
> On Thu, Jul 1, 2010 at 11:02 AM, Dearlove, Christopher (UK)
> <[e-mail]
> <mailto:[e-mail]>
> <mailto:[e-mail]
> <mailto:[e-mail]>>>
>
> wrote:
>
> Charles Perkins
>> I'd be happy if it were possible for [autoconf] to be allowed to
>> consider the excellent body of work that was seen already years ago
>> -- the same body of work that motivated me and others to create and
>> spend a lot of time over the last years and years.
>
> If Charlie can find a few like-minded people to work on that, why not
> add this as a parallel activity? The rationale of why two cases
> should be straightforward to make, they are almost chalk and cheese
> in e.g. centralised versus non-centralised. This is actually added
> safety to the group producing something, as if one succeeds and the
> other fails, that's still good.
>
>
>
> I also think this parallel approach could be appropriate too.
>
>
> Unfortunately, I can't offer to be one of those people. Although I
> should be able to contribute at the read and comment level, more than
> that is needed. I only didn't make the suggestion earlier as it needs
> people (with all due respect to Charlie, plural) to do the work.
>
>
>
> I'd be happy to help out on the matter. So I guess we have a plural
> + contributors ;)
>
> cheers,
>
> Emmanuel
>
>
>
> _______________________________________________ Autoconf mailing
> list [e-mail] <mailto:[e-mail]>
> https://www.ietf.org/mailman/listinfo/autoconf
>
>
>
>
>
> _______________________________________________ Autoconf mailing
> list [e-mail]
> https://www.ietf.org/mailman/listinfo/autoconf