On Sun, 13 Feb 2005 15:02:53 -0500 (EST)
[EMAIL PROTECTED] (Stuart Brorson) wrote:

> > The new functions are very very nice to use and make programs
> > better it seems to me.  And from a programmer point of view,
> > once you use them, you really don't want to go back... ever, really!
> 
> I'll belive you, but please don't forget about the user!  Users don't
> like to be on the mandatory upgrade treadmill.  Recall that that's one
> of the charges against Microsoft: they put you on a treadmill.  I
> think we should offer a better experiece for the user.  Anyway,
> open-source is supposed to be all about empowering the user, right?

I knew it :(  You won't believe how I was cringing when I hit the send
button a while ago.  I knew I should be mentioning the users, really!

> > But when you break compatibility with a previous library version
> > you should also bump the major or minor version of your program
> > (not just the revision #) and to let the user know when he tries to
> > compile, a header file should have something like:
> > 
> > #if !GTK_CHECK_VERSION(2,4,0)
> > #error Gtk 2.4 is required to compile this package.
> > #endif
> 
> How about if we use substitute code for older functions, like Ales did
> with GTK-1.2 vs. GTK-2.0?

Maybe, but the problem with this is that the new Gtk functions are
dramatically different.  It really is a whole new api and the decision
becomes do you want to maintain two different code bases... merged
or separate.  It's similar to the PCB Gtk vs Xt ideas that are being
kicked around.

A big effort designing for two apis could be followed by the risk that one of
them falls into decay over time because new developers will... well, maybe
these ideas should be floated on geda-user so users on the list can
speak up with their thoughts...

But wouldn't this be a perfect application for your install CD?  Keep
an old tarball on the website and say, "Install this if you don't have
Gtk 2.4" and then "If you want the latest but don't want to install
Gtk 2.4, here's your CD!"

Bill

Reply via email to