[GRASS5] Leaving

[Bernhard Reiter]

On Sun, Nov 23, 2003 at 11:17:11AM +0100, Thierry Laronde wrote:
> I think that the major advantage that people could advertise about
> Tcl/Tk is portability. I put it aside since I have more to say about
> that below.

The second major one is stability.  Tcl/Tk is very stable.
Also Tcl/Tk is Free Software.

> Maintainability-1? When one installs Tcl/Tk and see the dependencies and
> amount of time needed simply to compile... On the other side, Motif for
> example depends on X and that's all and Motif is a standard ( Motif X11
> Toolkit (industry standard GUI (IEEE 1295)).

As Glynn explained in another thread,
Motif is not a nice Free Software player.

> Maintainability-2? On the coder side. One advantage is to obtain a
> result rapidly. Tcl/Tk is good for prototyping or for small applications
> when you don't want to spend much more time on designing an interface
> than on developing the core routines.
> But the time "gained" initially is several times lost if the interface
> is complex. These languages are not as good as more raw ones to
> "rectify" the "need curve" (to approach as closely as possible a weird
> need with small segments of code). Because the learning time has been
> smaller, the coders understand less of the internals resulting in code
> "not doing" what it was supposed to do.

I cannot judge this efficiently.
Some studies and major application in Tcl/Tk suggest
that it is a power and easy enough language.
Personally I prefer Python by a wide margin and never leared Tcl/Tk.

> Maintainability-3? Almost one half of the compiling/running problems
> reported are due to the change in Tcl/Tk (8.3 -> 8.4).
> The less dependencies you have, the most stability you can achieve.

The latter sentence is a general truth.
If you use it, you have to see the other side, too:
There is a tendency to reinvent the wheel
if you don't use components.
I agree that the Tcl/Tk maintainers could be more careful
with some of their language changes.

> Last but not least: portability? People have envisaged portability as:
> one has a X machine with a Y OS, how can I make the software run on
> (X,Y)? But there are revolutions (linked): free OSes and network.
> 
> [A discussion with Michael Barton some days ago has been instructive on
> this side]
> The portability problem is now: one has a X machine and STOP! We can
> provide an OS running on X. And that's all. This OS has not to be
> installed even on the local storage of the machine (disk) because it can
> be launched from a CD (see Knoppix) or even more importantly downloaded
> from the network.
> If you have a small application, used a couple of minutes every hour,
> dual booting is not an option. With a complete work environment like
> GRASS, this is not a problem.

I disagree here.
My company created a Free Software version of Knoppix (the original 
containts proprietary software). It is nice to experiment with stuff,
but basically useless on real workplaces. 
For once there is the costs in speed, next you get the problem
of getting your data in and out, including your configuration.

The most important point though is, that people spend a lot
of perceived learning time in front of their windowing system.
They don't feel okay when they have to change it
and won't be happy.

> The future, for GRASS as for others, is a network of workstation,
> accessing SAN and able to be used as a cluster for heavy computing. And
> who says network says, too, yes, thin clients. That is X and light
> weight (have mercy for the servers!). 
> So Xt... but a "leettle" simpler would be best: Gtk, Qt? There
> is Motif. So let's go for Motif.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 189 bytes
Desc: not available
Url : http://lists.osgeo.org/pipermail/grass-dev/attachments/20031124/c0347862/attachment.bin