Hacker News Reader

> not a big company piling gobs of money into a full-time team

Think this is a bit of a straw man, to be honest. I've done turns at AWS, Microsoft, Dropbox, and now a quant fund. Most of the big-impact projects at all of these places have been bottom-up initiatives, not top-down -- they've been things started by one or a couple people, wedged into spare time, because the existing ecosystem had gaps. Even things that turn into full-fledged, public-facing services usually start quite small. A former teammate at AWS was part of the team who launched Mac on EC2 -- he was the third engineer on the team, and he started on the project as a part-time engineer on loan from another team. In my experience, this is how things are usually built -- even for things that go on to be part of the re:Invent keynote.

Maybe for some software, but rarely for anything involving hardware unless you're dummying it up on an Arduino. The hardware platforms I've worked would handily incinerate my entire household net worth in a year or so and relentlessly consume full time attention.

Even a part time engineer on loan is probably more time than some dude's weekend.

But i think part of it is how fast things can be scaled up. Corporations can go from the tiny to large scale fast of things look good. It takes a lot of time for open source to build momentum.

Freedom as in Beer is starting to mean a lot more to people now. That could be considered a major feature at this point. I started telling people about my jellyfin setup and suddenly "normie" friends that had no obsession with archiving and self deployment like I do are asking for help setting up their NAS and finding cool things to deploy in https://github.com/awesome-selfhosted/awesome-selfhosted

This has always been the end game for me of open source, beyond all the other benefits: offer tools so good for free that when someone tries to charge people for it, they'll be met with a resounding "why would I pay you for that when I can get this for free?"

Not to mention open source software basically reigns supreme in Venezuela, India, Philippines, other places without high capital in global currencies.

> You usually only get that initial rush in open source when it's a big company driving the project in the first place (LLVM, say).

LLVM began as a research platform at UIUC. In terms of public record, Apple only began putting money into it around 2005 (primarily by hiring Lattner). It might have received other corporate funding before and during that time that isn’t well documented, but I don’t think it’s accurate to say that it was driven by a big company from the get-go.

Yes, and if it wasn't for that, it would still be research platform like many others that predated it in compiler development, like Amsterdam Compiler Toolkit for one such example.

No disagreement there. My point was just that LLVM is not an "initial rush" case, to use the parent's language. It's a success story for the "small group of people" model.

Did the rush come before or after the huge (I imagine millions per year for a team of Apple devs working on what was to become a core technology pillar) infusion of cash and warm bodies? It was a fairly quiet and incremental research project and then Apple zoomed it to the stars (and Clang wasn't always even guaranteed to be open sourced: that came later, two years after the Clang project started in 2005).

Of course, this is actually one of the more unique cases, because it was both well-funded and open-sourced in the end, and, though developed rapidly, is actually a decent project, technically. As opposed to the kind of behind-closed-door too-embarrassed-to-publish shitware that's probably powering things like Hives.

All the examples where FOSS is ahead, is in tools by and for programmers. They're acceptable, maybe with an advantage due to cost or lack of cloud subscription stuff, but otherwise not great for the rest.

Most everything I like about Linux seems to be driven by the red hat types more than the general community.

Home automation software is pretty trivial, HW is what matters, and until very recently the HW was tied to the proprietary software. Plus the software can't rely on you being the product like commercial stuff, so you have to pay for easy remote access on home assistant.

The other issue is even though the functionality is simple, two important features are good UI, and not randomly breaking on updates, which FOSS isn't great at.

I use FreeCAD, KDEnlive, and others all the time, because the commercial stuff just isn't practical, if you're not making 40k a year or having a company that is tech focused enough to have a license.

I'm glad it exists, I'll continue to contribute in GitHub if I see an easy to fix bug..., I'll probably keep using my DIY automation system because of a few specific features, But I'd really rather FOSS do stuff a tiny bit more like commercial does.

What is meant by 'a bit "FC"' in this context?

Frantically-cobbled: from April 3 2005, announced April 6 and becoming self-hosting on April 7 by Linus Torvalds to replace BitKeeper after they dumped Linux (the public announcement of BitKeeper scrapping the free version was April 5, but the writing was on the wall for a month or two before that).

Evidence of the rapid initial development can still be seen in the famously...quirky...CLI syntax.

> If you want examples of open source lagging the commercial equivalent...

In all these examples, the missing part is not the technical prowess, but the optimized math which makes these software useful, which is a serious secret sauce.

Evaluating numerical calculations correctly under all circumstances is a kind of dark art.

Develop better algorithms, and you'll be in the business of disrupting them. See, OpenFOAM.

Disclosure: I did my Ph.D. on this subject.

I disagree. Having tried most of the products on that list and their competitors, the problem isn't the math. They generally get that right enough for most applications.

The big problem is the user experience. FreeCAD has single-handedly soured me on 3D modeling in general because it is awful to work with. There are a dozen incompatible ways to do anything, and they all have weird undocumented limitations. And rather than getting basic 3D modeling working properly, they add a whole architecture workbench.

We saw this with Blender a few years ago. The thing which made it go from "toy product" to "viable commercial alternative" was a huge UI overhaul in 2.7x. Something similar happened in KiCad 6. Both Blender and KiCad have attracted significant commercial backing since then and are slowly starting to become serious alternatives to their proprietary competitors.

100% agree, but how do we actually make that happen? I've joined RealThunder's (major FreeCAD dev's) patreon, but other than that I'm just resigned to a long wait.

The two I’ve seen is 1. a company needs it and doesn’t care so they just make a huge investment in the tooling that they open source or 2. someone with a lot of free time/independently wealthy just puts their head down and goes at it.

Get someone who cares to throw their weight behind it. You need a way to get developers to a) care about it, and b) do a lot of non-fun work. A Patreon isn't going to work because then the devs just end up doing more of the things they enjoy.

It seems like getting a commercial company behind it is pretty much the only way to do it. You just can't tell an independent developer that the UI of their pet project sucks and they need to spend six months rewriting it from scratch. Heck, getting them to agree that the UI sucks is pretty much impossible already! Stuff like this is a lot easier when your primary incentive is "being paid a salary" rather than "having fun with my pet project".