[Widex] suggestions re timestamps

[Dave Raggett <[e-mail]>]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

In the face to face meeting on Thursday, we discussed the potential
problems with how REX specifies timestamps for updates that are
intended to be applied at a future time.

REX currently specifies these in terms of the renderer's clock as 
milliseconds since the start of 1970 GMT. The problem is the 
likelihood of a significant clock skew between the renderer and the 
server, and the latency incurred by the protocol. This makes it 
difficult to the server to set the appropriate timestamps when the 
update is to be applied in a relatively short time interval into the 
future.

In principle, the server may be able to estimate the clock skew
and the network latency from a sequence of interactions with the
renderer, but this is tricky and subject to uncertainties.

W3C's SMIL language deals with this by allowing you to specify
event timestamps in terms off timed offsets from other events
(the beginning or end of a give presentation). SMIL isn't itself
a solution to Widex as it doesn't deal with the means to transport
DOM events and updates.

Widex/REX could use the SMIL event-value syntax, which includes
a reference to a named event and a signed time offset. This
would be very general, but comes with an implementation overhead
that may not be acceptable in all environments. A simpler approach
would be use non-negative timed offsets from an implicit reference
event. You would be able to designate that a given event is the new
reference point, replacing the previous reference point. The initial 
reference point is the start of the session.

This technique avoids the need to record the times of events
as they are applied. You only need to record the time of the
current reference point. The syntax for identifying an event as
a reference point would involve a new attribute. The existing
timestamp would be refined as a non-negative integer denoting the
number of milliseconds from the current reference point.

For multimedia presentations, the audio track is often used as
the reference timebase since human perception finds disruption
to the audio more disturbing than to the visual presentation.
A question for the W3C REX task force is whether they are interested
in synchronizing SVG updates to streaming audio, and if so whether
the presentation can be paused/resumed. If yes, then this should
be taken into account in the semantics for timestamps, e.g. the
clock should be paused along with the presentation.

Comments?


  Dave Raggett <[e-mail]>  W3C lead for multimodal interaction
  http://www.w3.org/People/Raggett [phone] (or 867351)

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.1 (GNU/Linux)

iD8DBQFEJCkFb3AdEmxAsUsRAmG1AKCiQWEkIYxLYmAOhQIYlxmhnByyEgCgn3bl
+M+Bs9+lJX/0L5mtiu2xYvM=
=UZF3
-----END PGP SIGNATURE-----

_______________________________________________
Widex mailing list
[e-mail]
https://www1.ietf.org/mailman/listinfo/widex