[patch] wireless patch, take 2

[Robert Love]

On Thu, 2004-05-27 at 00:32 +0200, David Zeuthen wrote:

> Thinking a bit more about it, there's some nasty corner cases here like
> the fact that net.ethernet.80211.ess_id is updated before the change is
> complete; what happens if we cannot connect to that network etc. etc.

Joe and I discussed this last night over "a few" beers.

Your point was basically my contention.  I totally abhor the idea that
there is a disconnect between the property and reality.  Hate it hate it
hate it.  I thought of the same things: what if the ESSID change fails?
What about the race with the tick resetting the ESSID?  Or what if the
user just does not have the proper callouts in place?

In short, ugh.

> Maybe it would be even cleaner to have a directory
> 
>  /etc/hal/methods.d/<name_of_method>
>                     ..
> 
> and have the method org.freedesktop.Hal.Device.<name_of_method> execute
> the appropriate script? If non-zero is returned then an exception is
> thrown at the caller with the message from stderr of the script. Or
> something.
> 
> We'd still have to specify all methods in the spec and provide
> convenience functions in libhal but at least this gives an expandable
> framework for such tasks as setting the ESSID - I imagine we run into a
> few others like this in the future.
> 
> How does that sound?

This would work, actually.  But do we want the complexity?  Last night
Joe suggested to me just sticking the essid changing code in HAL.  That
works.  I know there was contention to that earlier, and I agree with
those points.  But from my point of view, the only doable solution is
that or no ESSID setting code in HAL at all.

	Robert Love



_______________________________________________
hal mailing list
hal at freedesktop.org
http://freedesktop.org/mailman/listinfo/hal