Re: oaf async activation

[Maciej Stachowiak <mjs eazel com>]

Dan Winship <danw helixcode com> writes:

> 
> Uh, no, it's not interesting actually. We used threads because it was
> easier to move our large existing code base over to that model. For
> Nautilus, I don't think there was much pre-existing code with a strong
> anti-async bias.
> 

Bonobo?

gnome-vfs? <duck>

In all seriousness though, I think in retrospect it might have been
easier to just use the gnome-vfs sync API and put the threading at the
Nautilus level than to fix the threading at the gnome-vfs level. We're
currently seeing a lot of performance problems that can be traced
directly back to the main loop based architecture (Pavel has the
profiler numbers to back this up). Further, all the async framework
stuff makes adding even the most trivial new gnome-vfs file operation
a multi-day excercise in pain.

I suspect we've faced most of the same problems that would have made
it hard for you guys to convert camel to be async.

In any case, I suspect it would be useful to share the experiences
Nautilus and Evolution have had with the different approaches before
either project makes a final decision to switch.

 - Maciej