linuxaudio.org

On Wednesday 14 November 2007 02:49, Mark Constable wrote:

quoted text
> Audacity has it's aup format files that could be useful.

The .aup format was the first thing I looked at.  Audacity itself is

not scriptable enough to use in a headless web application, but its

format seems to be a pretty straightforward XML EDL.
> No reason why anyone can't "fork" the current piece and start a

quoted text
> new one from a particular point of the old ones evolvement.

Right, and while forking is usually a bad sign in the free software

field, I would hope the culture that springs up around the kind of

tool we're talking about would see it in a more positive light.
> Bingo, I've been looking around for a torrent based server/client

quoted text
> system that handles revisions and delta updates but it doesn't

> exist.

> T'would be a great project for anyone into deeper coding. I'd

> go for Git as the revision backend with a C/C++ torrent lib OR

> perhaps a distributed rsync client/server system.

I don't know that any of CVS, SVN or Git will be suitable for what

we're doing here.  Keep in mind that this system would most likely

never be receiving any patches at all, just new tracks and EDLs.

It's similar in concept, but still something that hasn't really been

tried before as far as I know.
On Wednesday 14 November 2007 02:59, Mark Constable wrote:

quoted text
> Perhaps a judicious compromise between using MIDI, where

> feasible, and vorbis test tracks (again, where feasible) and

> only relying on flac'd wav tracks towards the end of the

> pieces lifetime when it comes time for mastering.

That works great when everyone's purely an electronic musician, and I

participated in some "email GM files across the Atlantic" projects in

the mid-90's, but I personally enjoy using real instruments along

with all the samples and virtual synths.  Often I start with a guitar

or bass riff, and in those cases it's going to involve a waveform

from the very first track I lay down.  
If I compose a melody by humming it or playing it on the piano or

something, sure, it'll start out as a MIDI file (or XM module, if all

I have on me is my Nintendo DS, or in the past a Buzz machine or

whatever) but I don't just compose in a single way.  Having listened

to some other people's posted songs, I have a feeling I'm not alone

in this.
In any case, I think using any existing binary diff tool would be

basically impossible for a structured file like a MIDI file, and

difficult at best for FLAC files.  Again we'd end up needing to write

something new that would be of limited purpose outside our community.

I think if someone wants to upload a new fragment of a track,

a "patch" if you will, he or she needs to just upload that segment as

a new track and the software would let them cut it into the track

they're "patching", again using an EDL.
We're already talking about requiring FLAC uploads and that means

using the native file formats of Audacity, Ardour, Aldrin or any

other commonly used tool in a traditional CVS/SVN environment is

probably not going to work; it'd be up to the contributors to

maintain the original guide track they're working from (click track,

previous mixdown, or whatever) and export a FLAC track that they

could upload to the project and then tell the web app where it needs

to go (its offset, I guess.)
Rob