Re: [Autoconf] Working Group Last Call for draft-ietf-autoconf-addr-model-01

Did this mail reach any destination?
Teco.

>-----Oorspronkelijk bericht-----
>Van: Teco Boot [mailto:teco@inf-net.nl]
>Verzonden: maandag 4 januari 2010 20:08
>Aan: 'Thomas Heide Clausen'
>CC: '<autoconf@ietf.org>'
>Onderwerp: RE: [Autoconf] Working Group Last Call for draft-ietf-
>autoconf-addr-model-01
>
>Hi Thomas,
>
>To begin with: happy new year.
>
>And on your question, I expect the doc editors carefully check
>WGLC comments and update the doc.
>
>On this particular issue on 1-hop connectivity, the doc has the
>following text:
>>  If L2 communication is enabled between a pair of interfaces, IP
>>  packet exchange is enabled regardless of the IP subnet configuration
>>  on each of these interfaces.
>This is exactly my requirement. But for IP packet exchange, there
>is a need for a routing table entry or neighbor cache entry for the
>destination. I dislike an addressing model where 1-hop connectivity
>dependents on a routing protocol. Meanwhile, it is perfectly right
>that a routing protocol overrides a routing table entry for a directly
>connected interface with a longer match entry, even when it is a
>multi-hop path.
>
>The mechanism I use decouples configured prefixes on the MANET
>interfaces and the advertised prefixes. With a default configuration,
>only the MANET interface address is advertised (e.g. a /32 prefix),
>even if a shorter subnet mask is configured on that interface. Today,
>it is not an autoconfigured configuration.
>
>The question is: do we want to support such functionality in Autoconf?
>This requires providing and configuring a subnet mask. For IPv6, this is
>not a big deal, we can use /64 and define a well known ULA or whatever.
>This is also proposed by others, e.g. draft-perkins-manet-autoconf:
>>  In case the node does not know any suitable prefix, it uses the
>>  MANET_PREFIX, with prefix length 64, reserved for this purpose.
>For IPv4, this draft suggests the 169.254/16, which was zero-config at
>that time. Now, it is clearly defined as link-local. Try to
>allocate 240/8? Or leave IPv4 for what it is? Other options?
>
>I hope this helps. Did it?
>
>Regards, Teco
>
>
>>-----Oorspronkelijk bericht-----
>>Van: Thomas Heide Clausen [mailto:ietf@thomasclausen.org]
>>Verzonden: zondag 27 december 2009 5:33
>>Aan: Teco Boot
>>CC: <autoconf@ietf.org>
>>Onderwerp: Re: [Autoconf] Working Group Last Call for draft-ietf-
>>autoconf-addr-model-01
>>
>>Teco,
>>
>>What are you, concretely, suggesting to do to the doc?
>>
>>Seasonal greetings,
>>
>>Thomas
>>
>>
>>On 24 Dec 2009, at 08:32, "Teco Boot" <teco@inf-net.nl> wrote:
>>
>>> I have to add the issue on network operations (see my mail below).
>>> For IPv6, a solution is using LLs. Not easy to handle, but it works.
>>> But for IPv4, I can't find something working. And I did not receive
>>> any response on my mail.
>>>
>>> Because I work on MANETs that are actually deployed, and those need
>>> remote management, including a fall-back reachability option in cases
>>> the MANET routing protocol is not running, I swapped back to the
>>> addressing model I used before, that is using a common prefix for all
>>> interfaces to a MANET segment. I did not detect any problems with it.
>>> I had problems with the /32: some of the boxes I use simply do not
>>> support this, and when configured, I miss the fallback reachability.
>>>
>>> No misunderstanding: I keep supporting the advertised /32 prefix,
>>> advertised prefixes shall not overlap. This needs a function in the
>>> MANET routing protocol, that is does not advertise the configured
>>> prefix, but instead the /32 route (or a configured shorter prefix).
>>> All MANET Routing protocol implementations I use support this
>>> function.
>>> Write down a standard, for what is widely deployed?
>>>
>>> Regards, Teco
>>>
>>>
>>> PS. I think Autoconf should work on IPv6. And the addressing model
>>> should work well for multi-homed MANETs. The current draft attempts
>>> to define the model for both IPv4 and IPv6 and does not address the
>>> multi-homed scenario at all.
>>>
>>>
>>> -----Oorspronkelijk bericht-----
>>> Van: autoconf-bounces@ietf.org [mailto:autoconf-bounces@ietf.org]
>>> Namens
>>> Teco Boot
>>> Verzonden: dinsdag 8 december 2009 15:41
>>> Aan: autoconf@ietf.org
>>> Onderwerp: [Autoconf] 1-hop reachability depending on MANET protocol
>>>
>>> When using the proposed addressing model, I faced a reachability
>>> problem between 1-hop neighbors, when the MANET routing protocol
>>> was stopped. I couldn't start via the network, because the problem.
>>>
>>> Luckily, there are link-locals.
>>>
>>> Teco.
>>>
>>>
>>>
>>> _______________________________________________
>>> Autoconf mailing list
>>> Autoconf@ietf.org
>>> https://www.ietf.org/mailman/listinfo/autoconf