Linux sampler projects, was Re: [linux-audio-user] Some music made with Linux

[Mark Knecht]

Jesse Chappell wrote:
> Benno Senoner wrote on Mon, 23-Feb-2004:
>  > Not true Mark :)
>  > LS does not have this limitation: although the current implementation 
>  > for now allows only
>  > ALSA MIDI input, Christian added a time stamped event system which was 
>  > planned anyway
>  >  for accuracy reasons and because it will allow LS to be turned into a 
>  > VSTi or AU plugin.
>  > If some VSTi like standard and sequencers will arise under Linux, LS can 
>  > support such features
>  > without needing many modifications to the engine.
>  
> In addition, work is being done on midi-over-JACK which will provide
> sample accurate timestamped midi events via a new jack port type.  Programs
> that use this interface will get all the benefits of sample-synced
> realtime performance from their sampler/synth, plus the ability
> to render accurately non-realtime when JACK is in "freewheel" mode.
> 
> jlc
>  
> 

Hey guys,
    I ain't arguin', but I'll believe it when I see it.

    Please explain to me how any of this is supposed to work for the big 
world out there using Windows and Mac sequencer platforms. Today we are 
*all* using GSt on separate boxes for our sequencers. I just don't see 
it. These platforms are going to generate a MIDI event. They don't know 
a GSt MIDI event from an LS MIDI event. LS is going to play the note 
when it receives the MIDI event. How can it do anything differently? 
MIDI itself doesn't send a timestamp. How's that supposed to work?

    Even in the case of the VSTi wrapper on the Windows platforms (as 
has been discussed) I cannot understand that even if you can timestamp a 
MIDI event before you send it over Ethernet (and I'm not sure you can 
accurately enough in those environments) how are you going to do this 
when there is jitter in the clocking on those systems?

    I'll submit that on a Linux only set of boxes (well, possibly 
plural) it may be very possible to do something non-standard and get it 
to work, although I'd like to now what the latency overhead is going to 
be. If that's what you're going for then that could be cool for Linux 
only users. (Not the goal ofthe LS project I thought, but I've certainly 
been wrong before.)

    Anyway, it sounds great to me, and I'm here to test when it starts 
becoming real, but I just don't see it yet.

- Mark