Re: [DNA-BOF] Attachment Detection charter text for discussion

[Alper Yegin <alper@docomolabs-usa.com>]

> I would say "to work out what has been attached to" may be a part of
> DNA. But I don't have a strong opinion on that.
> 
> Regarding the the post-attachment authorization phase, perhaps one
> approach would be to
> 
> (1) Keep this primarily a problem of the folks who are doing it, such
>    as PANA. They could document the API requirements and implications
>    of back-off procedures etc. Or maybe already have done that. Alper?

We don't have anything to that effect in the spec yet. Before PANA succeeds,
the host will have an "IP-enabled interface with limited access
authorization". An implementation can mark the interface with an appropriate
flag to indicate this state.

Btw, another ad-hoc scheme in addition to web-based login is the VPN-based
access control. In that, a host is allowed to freely associate with the WLAN
link and given an IP address. But it cannot send any useful data traffic
except for connecting to the VPN gateway to access the enterprise network.
This is another scenario to consider...

> 
> (2) In addition, as we describe the link state and hints in DNA,
>    we'll add some clarifying text about post-attachment state and
>    hints. For instance, we could say that IP capability may not imply
>    connectivity, and that there may be additional tasks before
>    global signaling such as MIPv6 can take place. Name a few examples,
>    such as the login webpages, and say that details of these are
>    out of scope for DNA.
> 
>    The purpose of this clarifying text is to alert the reader to
>    the issue, but not solve it in DNA.

If we can somewhat abstract these interim steps and present how they should
be accounted in DNA algorithm, this would be good.


> 
> (3) If we are going to talk about multihoming case, mention the
>    item 2 problem in that context as well. Might want to prefer
>    a slow link over fast but webpage-login-not-yet-done link ;-)

I'm not sure we should be dealing with "interface selection" in this WG. I'd
think we should look at DNA on a given interface, not for the whole system
(host).


Alper