David Megginson said:

> Curtis L. Olson wrote:
> 
> > The other thing to consider is everyone seems to have one more fix 
> > they'd like to get in.  If we waited for everyone to be happy, we 
> > literally would never be able to have a release.  At some point we have 
> > to draw the line and ship.  The 0.9.4 release is simply the current 
> > state of cvs as of today, so if more people tried the cvs version and 
> > made patches along the way, we'd have less surprises on release day.
> 
> I agree with Curt.  There are two basic strategies for releasing:
> 
> 1. Release rarely, testing every release exhaustively first.
> 
> 2. Release often, testing every release only lightly.
> 
> I think that #2 works better for most cases -- many bugs won't show up 
> during testing by the members of the developers' list anyway, so it's best 
> to get the release out into the wild and find the real problems earlier 
> rather than later.

At one time I think I would have leaned toward #1 but have since become a #2
fan.  I was going to say this yesterday,  but I also realize that doing #2
often involves quite a time commitment.  Would it make sense to run a nightly
script for folks that don't run cvs?  Or is that just a waste of bandwidth? 
Maybe binaries are the ticket.  How about various team members running nightly
build scripts and then uploading the results somewhere?
 
> I know that some people like the idea of separate development and stable 
> branches, but unless we're dealing with critical core infrastructure like a 
> kernel or Web server, it's hard to motivate people to spend all the extra 
> time to backport bug fixes and OS compability changes to the stable branch: 
> it often ends up that the development branch is more stable than the 
> so-called stable branch anyway, while making twice as much work for the 
> maintainers.

That's right.  Maybe some day simgear,  but I would think that backporting
efforts would be driven by the requirements of the backporters rather than
anticipated as a requirement.
 
> I would support a 4-week code freeze before 1.0, however, and I do think 
> it's just about time for a 1.0.
> 

We're close, maybe next release, but I think we need to get the
subsystem/initialization thing straightened out first.  Fred's fix for the
tilemanager last week raised an issue about something that might be missing in
the subsystem design.

Best,

Jim


_______________________________________________
Flightgear-devel mailing list
[EMAIL PROTECTED]
http://mail.flightgear.org/mailman/listinfo/flightgear-devel

Reply via email to