>On Mon, 2004-08-30 at 16:56, Julian Rohrhuber wrote:
>> osc has quite a good system for timing I think.
>> It uses the NTP time server, to create time stamps according to the
>> osc time format (64 bit, first 32bit is seconds since 1970, second 32
>> is divisions of a second).
>> You then simply schedule the packets ahead in time, so the maximum
>> lag time is accounted for. This way you get very exact timing.
>
>Yep but OSC doesn't really have a system for timing beyond being able to
>specify that timestamp. I mean, OSC doesn't directly interact with NTP
>beyond borrowing its timestamp format, does it?
>
>http://www.cnmat.berkeley.edu/OpenSoundControl/OSC-spec.html
>
>NTP is great for synchronising clocks before and during a live jam, but
>the various systems also need to start at the same time and go at the
>same speed. If one joins in while others are running then it needs to
>be able to work out when to start, and when the speed is to change all
>the systems need to know about it.
yes, this is the funny thing - time stays relative and interwined
with space, whatever we try..
On the osc list there was expressed the need for a synching method
for OSC, so that on a request anyone can join in. I think that would
be the best standard, because once you have synchronized clocks, the
rest is easy.
>
>It would be great to work out a TOPLAP standard for time synch, a simple
>one that is easy to implement. Does anyone feel like working something
>out together? I don't mind taking my system apart and describing it in
>detail if it would help, but bear in mind that while it works in
>practice without drifting, it is fairly naive.
we could agree on an osc message that contains a tempo information,
maybe even, as a start, a simple clock tick without a timestamp, as
LANs are quite fast anyway.
>The time server is here:
>http://cpan.org/authors/id/Y/YA/YAXU/perl-music-article/examples/tm-0.1.pl
>
>and the client side stuff is here, in the spread_event_loop() routine:
>http://cpan.org/authors/id/Y/YA/YAXU/perl-music-article/examples/feedback-0.1.pl
>
>I'm using the 'spread' protocol and libraries to send timing information
>around out of programmer's laziness, it would be better to use OSC for
>this. When it comes to sending messages to SuperCollider, I'm using
>timestamped OSC bundles, with a little bit of buffering as Julian
>describes.
>
>Or does an OSC based timing protocol already exist without me knowing
>about it? How does SuperCollider deal with this? Klippav, how do you
>keep SuperCollider and Max in synch?
>
>alex
--
.
Received on Mon Aug 30 2004 - 19:39:57 BST