Re: [Autoconf] Updated ad hoc addressing model document

["Teco Boot" <[e-mail]>]

Thomas,

Late response. I have little time to work on Autoconf. 
I have to deploy something. Sorry for that.


>-----Oorspronkelijk bericht-----
>Van: Thomas Heide Clausen [mailto:[e-mail]]
>Verzonden: vrijdag 5 februari 2010 18:49
>Aan: Teco Boot
>CC: [e-mail]
>Onderwerp: Re: [Autoconf] Updated ad hoc addressing model document
>
>Dear Teco,
>
>On Feb 5, 2010, at 17:36 PM, Teco Boot wrote:
>
>> Thomas,
>>
>> Is this catching words?
>
>Not at all. It was explaining what I meant by "something clever" --
>you, incorrectly, read something into that phrase that wasn't intended.
>
>>
>> Agreed that standard "not that clever" behavior of the IP stack is
>> putting prefixes, configured on interfaces, in the routing table?
>>
>> Agreed that putting other prefixes is 'something clever'?
>>
>> Agreed that with the proposed addressing model, under conditions that
>> the 'something clever' is not functioning, L3 communication fails for
>> links between two nodes that have L2 communication?
>
>No, I do not agree with the above.

I think you miss something important here. 


>> If I am missing something, please come up with it.
>
>That's not how it works.

I'm posting this issue AFTER testing.
And I'm quite familiar with currently shipped MANET products.
I can't see how these can support the suggested addressing model.
I also asked Mark on this. No response yet !!

Far more important. It doesn't work when I post during WGLC, and it is 
ignored totally. I checked mail archives, there wasn't a single response
form doc editors for clarification, nor corrections in the draft.


>There's a document that has attained what appears to be relatively
>wide consensus. 

OK, this is up to you as chair. But wide consensus?


>                You state that there's something wrong with the
>document and that you have an "alternative model". That's fair enough,
>but then the onus is on you to explain what is wrong, lay out your
>"alternative model", and make sure that in doing so it addresses also
>the concerns which draft-ietf-autoconf-adhoc-addr-model raise.

I explained what is wrong. And at least one (Carlos) understands
the issue in the document.

Again, my response during WGLC on this topic:
Page 4:
   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 only possible if the IP stack is aware of the L2 link between the 
pair of interfaces. One way of establishing this is configure an on-link 
subnet prefix. Now we choose not to configure so, and instead run a MANET
protocol. But this makes IP communication dependent on the MANET protocol, 
which can have a negative effect on managing the network. Think of remote
management and updates on the MANET protocol.


>Absent that (and we have had since the Stockholm IETF), I propose that
>we move forward draft-ietf-autoconf-adhoc-addr-model.

I might write something. No time today.


Thanks, Teco