Matlab is truly the RAD tool of choice for numerical programming and has a solid grip in universities combined with enterprise-level support.
I do not think Octave ever tried to replicate its workflow (which is not general programmer centric at all) and domain-specific documentation but merely focused on the underlying language compatibility, which is really the least important part of Matlab.
On top of that, I seem to recall, Matlab was one of the first of the specialist programming toolsets to offer a very competitive "Student Edition". This was a godsend for schools and universities before the Internet took off.
In short, Octave was too little too late, and Numpy/Scipy, while catching up fast, has supporting tools spread all over the place as well as being geared more to general programmers who want access to convenient numerics rather than numerical modellers/engineers wanting a RAD tool.
Numpy/Scipy etc. may well overtake Matlab eventually, but that will be purely a function of its infrastructure, not the something as mundane as even the nice language (which admittedly was its initial driving force). At least in this respect, it has done a lot better than Octave in much less time.
Actually, there are tons of people who want to run Matlab code freely. The code is already written. They need to run it in clusters, or they need to run it at home.
I do not think Octave ever tried to replicate its workflow (which is not general programmer centric at all) and domain-specific documentation but merely focused on the underlying language compatibility, which is really the least important part of Matlab.
On top of that, I seem to recall, Matlab was one of the first of the specialist programming toolsets to offer a very competitive "Student Edition". This was a godsend for schools and universities before the Internet took off.
In short, Octave was too little too late, and Numpy/Scipy, while catching up fast, has supporting tools spread all over the place as well as being geared more to general programmers who want access to convenient numerics rather than numerical modellers/engineers wanting a RAD tool.
Numpy/Scipy etc. may well overtake Matlab eventually, but that will be purely a function of its infrastructure, not the something as mundane as even the nice language (which admittedly was its initial driving force). At least in this respect, it has done a lot better than Octave in much less time.