[approach-discuss] first draft of Network Auto-Magic architecture

[John Beck]

> http://www.opensolaris.org/os/community/approachability/nwam/architecture/

> Draft 0.1, 2006-Feb-13

Revised to 0.1.1 and today.


David> Section 1, "At any given time, ..." - Does this mean that I would have
David> to duplicate punched-into-SWAN settings across child profiles layered
David> on top of each network that I may punch in from?

That was not intended, though I can see someone inferring that from the
current text.  We'll have to work on improving that.

David> Not to mention overlap with my directly-connected-to-SWAN settings.
David> Shouldn't layered configuration be independent of base configuration?

Yes; again, the inference you are drawing was not intended, though it is
perfectly reasonable for someone reading the current text to make such an
inference, so we will rework the text.  Suggestions for such rework are
always welcome and appreciated of course.  :-)


David> Section 3, bullet 6 - The State Machine section does not discuss
David> "bouncing" interface at all.

David> - "... link change is transient or note." should probably be "... not."

Renee responded to both of these, and I have added some clarifying text to
the bullet in section 3, and also fixed the typo.


David> Section 4, bullet 5 - What does "query always, query only if new" mean
David> in response to a new wireless AP?

The former means "always query the user as to whether or not to connect";
the latter means "query the user if this is a new WLAN, otherwise connect
without querying".  I have added some clarifying text, please let me know
(off-line if you wish) if you think it still needs work.

David> Shouldn't you distinguish between a new wireless AP with a new
David> network name and one with a known network name?

I'm not sure I follow you.  I run into WLANs with the name (essid) "netgear"
all the time, so it seems more info (perhaps bssid) would need to be kept
track of as well.  Perhaps you'd care to provide an example?

-- John