On Aug 23, 2010, at 16:17, Peter Bigot wrote:

> When I first read the HC drafts, I interpreted "context" to permit
> external knowledge, such as the fact that context 7 that I used it
> told me to construct the IPv6 address from a locally stored constant
> that was not part of any ND message.  Whether to pad with zeros or
> with FF:FE000 smells like this sort of external knowledge.

Generally, we have tried for each context to take the form of a prefix (a 
number of bits plus that number).
More complex contexts are possible, but we'd also like to be able to distribute 
them using 6lowpan-ND.

> Will it be required that padding be done this way for a 64-bit context
> and 16-bit short address, or merely allowed?  Is padding-with-zeroes
> as an equally valid interpretation of the context?  

Not for the contexts I'm talking about.

> Does this freedom
> to interpret apply to other context lengths where the prefix length
> plus the address length does not equal 128?
> 
> Can I say that context 7 in a network I control means ignore any
> provided prefix and use a hard-coded constant 112-bit prefix, if I
> have some reason to believe this makes sense in my deployment
> environment?

Probably, but then you can't distribute that context using 6lowpan-ND.
Or, in other words, you can only use nodes that already have that knowledge.

Gruesse, Carsten

_______________________________________________
6lowpan mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/6lowpan

Reply via email to