>>>>> "Carl" == Carl Ollivier-Gooch <[EMAIL PROTECTED]> writes:
Carl> Let's not forget here the possible (probable?) annoyance to
Carl> users of being regularly encouraged to upgrade to a new version.
Carl> I know that I won't personally upgrade (on five different
Carl> machines, counting my students' machines in the lab) until a new
Carl> can't-do-without feature comes along. Many users will probably
Carl> feel the same way, which will have two effects. First, the most
Carl> recent releases will not be thoroughly tested. Second, bug
Carl> reports may go back several releases at any given time. (Yes,
Carl> the two are connected.) While this isn't intrinsically a problem
Carl> from the development side, I suppose, it -does- mean regular
Carl> checkouts of old releases by developers to try and figure out
Carl> what went wrong with version 1.20.0 when 1.34.0 is current.
That's why using fetchmail numbering scheme would be nice: we work on
lyx 1.1.x releases uuntil we have both new features and stability, and
then release that as 1.2.0. This means that
- 1.x.0 version will not happen very often, so people will be
encouraged to upgrade
- all versions are (supposed to be) stable versions, so adventurous
users can try them out.
Moreover, cvs can tell us what changes have happened between 1.20.0
and 1.34.0. And if we cannot sort it out (or reproduce the problem),
we'll ask the user to upgrade first ;)
Seriously, this scheme will not be worse than what happens currently
with preversions between stable versions. We just dropped the notion
of a development branch, which needs much more core developpers than
what we have currently.
JMarc
