On Tue, Oct 17, 2006 at 06:04:54PM -0700, John Beck wrote:
>
> Sebastien> The NWAM design seems to place all configuration of data-links and
> Sebastien> IP interfaces as part of a LLP, and an IP tunnel is definitely
> Sebastien> both of those things. So, according to the design, an IP tunnel
> Sebastien> should be part of a LLP. At the same time, the NWAM architecture
> Sebastien> describes an upper-layer profile as something that defines a
> Sebastien> set of things that happen under certain conditions, such as the
> Sebastien> configuration of IP addresses in a particular subnet, or which
> Sebastien> physical links are connected. It would make sense to make the
> Sebastien> configuration of an IP tunnel conditional on such criteria.
>
> You are following our architecture/design correctly so far.
>
> Sebastien> As such, I don't see the LLP vs. ULP distinction as being very
> Sebastien> clear cut for IP tunnels, and I think I'd benefit from a whack on
> Sebastien> the head by a heavy clue stick. :-)
>
> No, you don't need a whack; you have pointed out a gray area which we have
> been trying to resolve. Stay tuned for more on this...
IP tunnels are tough to categorize. So far, I think we've assumed the
following attributes for LLPs and ULPs:
LLPs
1. Contain config info for Layer 3 and below
2. Contain attributes that apply to individual interfaces or links
ULPs
1. Contain config info for things that operate above Layer 3; or
more generally, for things that don't care so much about what
interfaces are there or will be used, but just want to send a
packet to IP for delivery to a particular destination
2. Contain system-wide attributes
3. Are selected based on current network availability
The problem is that (as you pointed out) IP tunnels match LLP items 1 and
2, but also ULP item 3. We definitely need to do more thinking about how
they should fit in.
-renee
