Peter Moulder wrote:
Later that day, Michael & sciboy did get a connection. Issues:
- Lots of output on the terminal, should be turned off for general use before release.
Whoops...I thought I killed all of the useless output in CVS. I'll have to go back and check that.
- Michael & sciboy's session got out of sync somehow: at some point sciboy cleared everything but a star was left on michael's session. Not a show-stopper: this was the only thing we were aware of being out of sync.
- Undo is global, i.e. it often undoes what the other person did rather than what the undoer did. This is a usability problem, but changing this is non-trivial, so I don't consider this a show-stopper.
The release notes contain notes to the effect that problems like this may occur. Future releases will contain a much improved architecture for handling concurrent modification, undo, and such. The model I'm currently studying for this is the Palimpsest model (http://cs-people.bu.edu/dgd/thesis/original_paper.html is a shortened version of the author's original dissertation).
Btw, Michael was using MacOS X, so it seems that libloudmouth works acceptably there (as well as on Linux/Gnu). I gather Windows still doesn't work acceptably well, due to threading differences.
Oh, cool :)
sciboy did experience a segfault later in his session with Michael (18:47 in the gristle log), though he didn't have a backtrace at the time. My feeling is that if the crash is specific to inkboard then we not consider it a show-stopper, but just make a note of it in the release notes.
Sure thing.
-- David