The problem with Octave for most of its life was it never had a very good GUI, and that made it hard for university departments to adopt as course software. I know we tried to push Octave in our department, but it just never took off. That, plus the Mathworks offered some pretty good educational pricing.

Octave also never had the span of toolboxes that MATLAB had. With MATLAB, you could always find fairly usable toolboxes that covered most engineering computation needs.

Octave was on par with MATLAB on some fronts, but it rarely exceeded MATLAB capabilities. And it was worse than MATLAB on tooling maturity. People hate The Mathworks for charging an arm and leg for its software, and for breaking backward compatibility, but most people will admit the MATLAB environment is pretty polished and seamless compared to open-source offerings in the same space.

So far no one has mentioned Scilab (http://www.scilab.org/). Scilab is a much a better MATLAB alternative than Octave, though Scilab code is less compatible with MATLAB that Octave code.

There's also FreeMat http://freemat.sourceforge.net/ which at the time had much better Matlab compatibility than Octave, and a much better GUI (comparable to Matlab's circa 2007 or so, which was when I compared them).

And for some reason, it was virtually unknown - Octave is superfamous compated to FreeMat.

Something you didn't mention, which I suspect is a major factor in why promoting FOSS alternatives proves surprisingly difficult in education.

I remember when I was in university students were acutely aware that tools like PhotoShop, Microsoft* and MATLAB were used in businesses. I doubt this has changed.

Even if superior Free options had been available, many of them would still have opted for the "safe" proprietary options because they wanted to be sure they could get work and felt (rightly or wrongly) that it was important they could put the right product names on their resumés.

I can't speak to Matlab/Octave but as for Photoshop, FOSS folks always say use GIMP. But the interface is just so terrible compared to PS and I think they underestimate how important that is. Maybe it's configurable but for example, simple things like alt-tabbing between GIMP and other programs was painful because it opened parts of its interface as separate windows, so you'd alt-tab to your browser, then back to GIMP, but it wouldn't show the artboard, just the tool panels. That's a separate alt-tab. And more generally, it felt like everything in GIMP took two extra clicks than the equivalent PS operation.

Most people are goal-oriented when it comes to using software, not ideological, and they'll go with what gets the job done with the least amount of effort.

What's funny is Gimp used to be single-window, then for some reason they redesigned to multi-window crap, and now they're back to single window (at the cost of ctrl+s -- fine, whatever). Still many UI quirks for sure. I don't know how much it matters for non-trained users though, Photoshop confuses me too on the rare occasion I get to try using it. For professional artists, they're likely using more tools than just "Photoshop for everything", so it's not like they're incapable of learning new UIs even if they suck. (Edit: And generally these days I'd rather advocate for more diversity in tooling, regardless of closed/openness, just to avoid monoliths and monoculture.)

The GIMP-over-Photoshop argument should be because it does something Photoshop doesn't (very possible with its ecosystem of plugins/filters -- for example a "smart remove" plugin was around for quite some time before Adobe made their own and spent who knows how much in marketing/demos for it) or because the artist wants to save money.

Agree that people are goal oriented. OSS that's free-as-in-beer can help drive adoption on account of being free, but ultimately the best OSS succeeds because it's better than its competition in important ways. Pure clones are risky, I'm glad Gimp doesn't try too much to be a clone.

It's not really about being incapable of learning new UIs though - I'm perfectly capable, I just don't want to. Why learn a UI that takes 3X as many steps to chop up a site design when there's one that takes X?

Sadly most artists who want to save money just pirate, which is what led to the terrible Creative Cloud situation. But that led to Affinity which seems to be a solid, buy-once (for cheap!) replacement. I agree the monolithic nature of Adobe is bad for everybody but that's not a reason to choose a subpar tool. Also people can write and sell their own plugins for Adobe products as well.

Anyway I think we're mostly on the same page.

You make a good point. MATLAB, rightly or wrongly, has become a signal for engineering computation skillset on resumes.

That said, the world has become more accommodating in other ways. Python and R are now more well-accepted signals for data skillset than say, SAS. (exceptions: finance, pharma)

While R is superficially very similar - having used both, my impression is that it's actually not for the same use-cases as Matlab/Octave. Matlab is primarily for linear algebra, while R is primarily for data-frame manipulation (which is more like SQL). So while they both can do the others job, things get clunky really fast

Out of curiosity, what are the pain points you run into using R for linear algebra? I've found it's not too bad for the basic things I've used it for, but my needs admittedly weren't very complex.