On being Matlab

[John W. Eaton]

[I moved this to the maintainers list. --jwe]

On  1-Nov-2006, Tom Holroyd wrote:

| I think I have had about enough. You know what? I suggest that 
| Matlab compatibility should NOT be a goal.

In that case, I think there is no point in being "sort of compatible"
so I say let's design a better language with cleaner syntax and
semantics.

| As I note above, I 
| don't believe it is actually possible. Octave is already close 
| enough, and the pain of switching from Matlab to Octave is 
| already no greater than the pain of upgrading Matlab itself.

The trouble with this is that since Matlab is a moving target, we will
soon be hit with lots of messages asking why Octave is missing so many
core features from Matlab (even if they really aren't core language
features).

| I think if people want to use Octave, they will. They will change 
| their code to the Octave way, and they will be happy if they 
| never have to change it again because of an upgrade.

I think the big increase in mailing list traffic is due precisely to
the fact that Octave *is* mostly compatible.  Didn't people want
Octave to gain users?

| I think it might even be helpful to take a step back, and not 
| worry about being compatible. Remove some of the compatibility 
| stuff, in fact.

Instead of doing this, I propose just picking off the easy targets,
and when there is something that is hard to do, simply saying we
won't fix it becuase it is a large job, so either someone who cares
will have to provide a patch (a good one) or sufficient funding to
motivate someone else to do it.  For this to be more useful, we
probably also need to start thinking about a bug tracker of some kind,
so we can classify bugs and archive them in some way, so when people
ask for a feature that has already been discussed, we can easily point
them to the discussion on the bug tracker web page.

| Don't warn people that their code might not work 
| in Matlab. I bet they really don't care. Make it friendlier, so 
| that the first thing a first time Octave user sees _isn't_ a 
| blizzard of warnings. Set the defaults for maximal quiet. Pick a 
| code base of old, dusty, reliable .m files and make sure that 
| Octave parses the lot with nary a whisper. Call it Octave 
| Language, lock the door, set for 3.0, and heave.
| 
| Well, you are still allowed to add new features, of course. :-)
| 
| Seriously, though, I think the first-time experience of a crusty 
| old Matlab user should not be a mess of warnings.

I wasn't aware that Octave was generating a lot of warnings about
Matlab incompatible code.  In the current sources, the
Octave:matlab-incompatible warning ID is set to "off" for both the
normal and --traditional modes.  So I don't understand why that is
happening for you.  It is not intentional.

| <flame off>

Foul!  I didn't see a <flame on>.  :-/

jwe