4 ms·
We've been thinking about this and working on it a lot lately. I hope to announce our efforts on HN soon (company is in stealth mode and isn't launched yet). T
by mike_hearn 5y ago
We've been thinking about this and working on it a lot lately. I hope to announce our efforts on HN soon (company is in stealth mode and isn't launched yet).
There's really many different aspects to this that you can't do justice to in an HN comment. But firstly let's observe that web vs non-web isn't deeply fundamental. Steve Jobs himself tried to push web apps on mobile devs and they rejected it. Many other people have tried to push mobile web apps and generally, failed. Users wanted apps built with non-web technologies.
So why hasn't this happened on the desktop? I think it's a combination of:
- No sandbox. In turn this means, IT departments are scared of desktop apps and have to whitelist them but many IT departments are bureaucratic and slow, so this can be painful. This is exacerbated by:
- Poor deployment technologies in general, with a poor understanding of what IT departments need. On MacOS X there's no software update system out of the box except the App Store, which comes with massive downsides. On Windows, IT depts generally need some sort of consistent package management system but Windows apps all invent their own, because up until recently with MSIX, Windows has not provided what was needed. Most apps are thus painful to (re)package and control. Witness the screams of Windows network admins about Electron - it's a disaster for them. In effect every Electron apps (using Squirrel) installs into roaming $HOME meaning when users roam massive apps get copied around, slowing down login. This is all wrong but the Electron distribution systems are unmaintained, so it never gets fixed.
- Cross platform desktop UI toolkits aren't that great IMO especially for devs who want to avoid C++. Your best option is JavaFX, but it's not well known and doesn't have a wealthy tech firm who can subsidize its development (any more). Still, the open source community around it has been growing and the tech/capabilities are very solid.
That said, the alternative is HTML5 which isn't a UI toolkit at all. So, a lot of this is comparing apples and oranges.
- Conflation of unrelated aspects, e.g. a lot of people associate desktop app with pay-per-upgrade. You can do that with desktop apps, but you don't have to. You can implement subscription pricing too, c.f. JetBrains or lots of mobile-first SaaS firms. You actually have a lot more flexibility around pricing because your costs are more or less fixed.
You could go on - the web is an evolved technology with a form of evolutionary robustness that takes a lot of work to even analyze, let alone out compete.
Nonetheless, I feel strongly that the software industry should:
a. Analyze why we like the web as a technology.
b. Use that analysis to do better than it.
Here's a controversial opinion for you: HTML5 is reaching end-of-life as a technology. Google lavishly funds Chrome and they don't seem to know what to do with that money beyond try and turn HTML5 into a spec for an entire operating system, except it's one that ignores the entire history of OS design. The sheer size of the spec means it's become unimplementable! Not even Microsoft or Apple can fully implement it anymore meaning it's more or less Google's plaything. The openness was one of the big reasons to like the web; IMO that's no longer a competitive advantage especially as many competing platforms e.g. JVM+JavaFX are themselves open source and actually have multi-vendor communities around them, which Blink does not.
Meanwhile, despite the vast size of the spec, the platform itself doesn't really meet people's needs and hasn't innovated on the core tech. It's stuck in a local minima which kills progress. As a consequence many websites these days are layering many pre-processors and alternative languages on top that compile to the "machine code" of HTML/CSS/JS. Trivial example: the popularity of Markdown is more or less driven by the failure of HTML to be an ideal appearance-neutral markup format. Yet, browsers can't render Markdown sites natively.
We need new projects that try to build post-web technologies, ones that take the ideas that work (markup, sandboxing, a portable browser etc) and throws out or upgrades the ideas that aren't working.