mailboxes and telephones in the Web

It's fairly well documented that mailto:
is a misnomer; it should have been
mailbox: . (unfortunately, not well
documented enough to prevent [callto]
sigh...)

But recently in this forum[1,2] folks have talked
about browsers "representing" mailboxes
with compose windows, and being
able to click on a tel: link and
have the user agent dial the phone.

I suggest a slightly different way of
looking at it:

I implemented an HTTP proxy[telagent] that handles
tel: URIs by redirecting to a form
for "composing" a telephone call...
you compose the call in an HTML form,
and you POST to a dialer-resource.

Dialing a telephone is not something
I consider safe... not a GET sorta thing.

By analogy, I don't think web user
agents represent mailboxes by
compose windows; if you want to
think of them doing a GET on the mailto:
URI, I suggest the reply is a redirection
to a compose window.


I think one of the best ways to test
an architectural principle is
the "when is the next time this
matters?" test. And I think integrating
telephones (and IP telephony, and
video, and instant-messaging,
and chat and the like) is one
of the next times that the
architectural principles around
naming and REST apply.

I found it straightforward to apply
the existing GET/PUT/POST model
to telephones. Does it seem natural
to other folks? or was I stretching it?

Is this a useful scenario to include
in the naming or REST section somewhere?


[callto] http://www.w3.org/Addressing/schemes#callto

[1,2] I spent about 15 minutes
surfing for a reference; rather than
risking not sending this at all, I'll
just send it now and fill in the reference
later, if it's necessary and nobody
beats me to it.

[telagent] http://dev.w3.org/cvsweb/2001/telagent/

-- 
Dan Connolly, W3C http://www.w3.org/People/Connolly/

Received on Tuesday, 19 March 2002 17:32:13 UTC