[GRASS5] Leaving

[Glynn Clements]

Markus Neteler wrote:

> > So the best method to see what is indeed the correct way is to test: I
> > will go my own way starting from the last CERL public release, since my
> > needs are not cutting edge ones and I have already done a fair amount 
> > of cleaning (this means mainly deletions, and 
> > standardisation/ansification is on the way ---and I have now a more 
> > clear idea of the work done by Markus since, yes, one can not say that 
> > if one wants to compile on a POSIX system with -Wall -Werror it does 
> > compile out of the box!).
> 
> Mhhh, time to fork the project? Maybe some developers would join you...
> (recalling the license discussion). I agree with you that more radical
> changes are needed, as already posted several times.
> Maybe LGPL'ed libraries with GPL'ed application layer? Well, most
> developers have posted their opinion on this already.

It's not clear to me exactly what Thierry's complaints with the
current direction are, but they don't appear to be related to
GRASS' licensing policy.

As for cleaning up, there's already a consensus that it needs to be
done; it just doesn't appear to be happening. Eliminating warnings and
converting to use ANSI prototypes are both tasks which can reasonably
be done in 5.3. [However, as I've said before, warnings shouldn't
simply be "made to go away". The wrong fix is worse than no fix, as it
almost completely eliminates the possibility of someone subsequently
making the right fix.]

Re-formatting is a bit more problematic, as it eliminates the
possibility of obtaining meaningful diffs between pre- and post-
versions. However, this isn't that critical, so long as "real" changes
aren't committed at the same time.

The one thing which shouldn't be done in 5.3 is to move or rename
files, as that resets the history. That doesn't apply to deleting
clones; if we delete a file, we no longer care about its history.

> > My plan are, too, to resurrect the X Window/Motif interface which is far
> > more challenging, at least for me: I know at the moment strictly nothing
> > about Motif programming and there has been a fair amount of changes
> > since 1991...
> 
> If Motif is a good choice, I am not sure. Looks a bit dated.

I don't think that Motif is a good choice here. And I seem to be one
of the few open-source programmers who hasn't succumbed to the Gtk/Qt
groupthink. I still think that Motif is a good choice for an
X-specific toolkit on technical grounds (in terms of acceptability, it
depends upon the target platform; Motif is a good choice on commercial
Unices, a poor one on Linux).

However, it definitely wouldn't be the right approach from a
portability perspective. Particularly given the licensing issues;
Motif is free for open-source platforms, but that doesn't apply to
MacOSX, and probably not to Cygwin (Cygwin is open-source, Windows
isn't). [Licensing isn't an issue for commercial Unices, as they
invariably include Motif with the OS.]

wxWindows has reasonable portability. Its use of C++ presents some
issues regarding binary compatibility, but then the same is true for
using Tcl/Tk.

FWIW, I have managed to get non-trivial Gtk programs working with the
Windows port of Gtk, which doesn't require Cygwin or an X server. I
presume that Gtk works fine on MacOSX with an X server.

-- 
Glynn Clements <glynn.clements at virgin.net>