This is probably the kind of person you don't want to cut ties with open source too often. Good software, from research to implementation takes a huge amount of resources (both people-hours and people to put in those hours), and the kind that measure 20 times before making 1 cut are some of the most important, because they are the type that should be in charge of the types of early design decisions.

For every one of these kinds of developers, it seems like there are at least 10 who measure way less and cut way more (myself included in that 10).

I do think ticki might find his way back to F/OSS with Haskell or some similarly related ML project, they can offer a good mix of ivory-tower and just enough in-the-weeds implementation to make rigid, type-safe, theoretically sound systems and abstractions...

Maybe it was just an issue of picking the right language to pursue the purity he was seeking -- one further away from the mainstream, under less pressure to "get shit done".

jstimpfle

In the first place, good software is software that exists. I don't know ticki or his work, but if it is vaporware what he calls vaporware, then there is no way for it to be good software.

If ticki reads this: props for the clear words. Honest posts like this are rarely seen. And good luck for your career in mathematics. Though I figure great mathematics is not much different than great software (I'm not talking about glue-coding): It needs to exist. You need to find simple, obviously correct, and useful models to create something that other things can build on.

And if you decide to come back later: A clear goal helps separating the essential from the non-essential. Formalism is only a tool, not an end in itself.

hardwaresofton

I don't agree with the notion that software has to be finished to be good (which isn't exactly what you said but the implication seems to be there). Taken literally, software that is unfinished (i.e. unusable) cannot possibly be good, obviously, but software that was been theorized but didn't necessarily exist in finished form until relatively recently is pretty common. If you don't want to make a bundled mess of spaghetti design, theory/play/experimentation that doesn't necessarily turn into a functioning project is crucial, ideas are sometimes reused/refactored into a project that is completed.

Maybe if none of his work was actually published at all, then that statement would make sense (since it would literally be impossible to derive value from something that doesn't exist), but he's contributed documentation, schematics, etc, it's vaporware in the sense that it isn't a finished product and may never actually become one since he's not supporting it anymore, but if someone else sees what he's done (ex. https://github.com/redox-os/tfs), they might pick it up and finish it. ATM there are 500+ commits to that project, I find it hard to believe that if what he was doing is worth doing that those commits are completely worthless.