The maintenance effort required on iOS is substantial. About a quarter of your full-time year needs to be dedicated to it.
On desktop, you can just publish your software and slowly see it age as you work on your next big release. On iOS, it ages every year at brutal pace, and your new sales will plummet while you work on your next big release, meaning your revenue crashes much faster.
Even worse, the iOS App Store has no notion of paid upgrades, and publishing a new app is basically like starting from scratch as far as discoverability goes. So when you finally have your next big release ready, it's like launching a completely new company.
Apple really wants developers to make subscription apps that ship frequent iterative changes, and other business models just simply don't work well on their mobile platform (on Android it's even worse btw).
Because untrustworthy websites can piggyback on the brand name.
"Download ffmpeg here: sudo bash -c ..."
And then the installation script from our malicious site installs ffmpeg just fine, plus some stuff you have no idea about. And you never know that you've just been hacked.
Can you repeat this mental exercise for every other installation method you can think of? e.g. distributing deb/rpm files, distributing AppImages, asking users to add your custom repository and signing key?
(Yes I know that the last one has built-in benefits for automatic updates but that's not going to protect you on initial installation and its benefits can be replicated in a more portable way in any other auto-update mechanism with a similar amount of effort)
((And if you have the patience to set up a custom repository, you can simplify initial installation process using a "curl|bash" script))
If you get your install instructions from an untrustworthy website, there’s nothing preventing them from telling you to use a third-party apt repository or ppa that gives you a malicious version of the thing.
There’s not really a difference between curl piped to bash, and installing packages from a third-party package repository that the distro maintainers have no involvement in with.
Edit: just tried it for serving a fastapi. It's fantastic. Instant TLS via Let's Encrypt. There may be other webservers that are equally easy, but this one is certainly easier than Apache or ngninx, which I used so far. Love it.
--
Reach out to the guys at Kamal. They wrote their own reverse proxy because they thought Traefik was too complex, but they might be super happy about yours if Ferron is more powerful yet easy to configure because it might solve more of Kamal’s problems.
I previously founded and sold an AI startup to Spotify; that doesn't actually make me smarter than the average HN user (mostly just more lucky) but it probably looks nice on a social proof section.
I finally started migrating our app from Heroku to Hetzner. Started by moving Postgres, and boy, what a difference in performance. I pay 1/5th of what I previously paid, and performance is through the roof. Like, as if we had rewritten our Rails app in Rust (metaphorically speaking). I know that in theory you get much more from managed postgres than a simple self-managed server, but I have pgbouncer and Wal-G set up, and it really seems like it's not that hard to manage that instance. Most of all, we were really hitting performance limits on Heroku, not theoretical "what if the server goes down" but real ones that users could tell every day.
Will move the app server itself next (basic Rails). Really wish I had done the move earlier. Hetzner is like cheating for indie devs.
Why for now? I see zero reason why that should change.
It's not like Heroku was zero maintenance. I often ran into resource limits. I did a fair amount of optimisation on jemalloc to make sure Rails didn't run into Heroku's ridiculously low RAM limits because of loading too many ActiveRecord objects. Yeah, sure, I should probably paginate better, but now I have so much RAM it literally doesn't matter.
Heroku also regularly asked me to schedule DB maintenance. Before I paid for the expensive DB, that meant the connection string could also change, which meant I had to redeploy my AI workers (i.e. background jobs that need a GPU) that were hosted outside of Heroku. Now it still warns me that the connection string would change, but it somehow didn't anymore. Ah well. All problems of the past.
The solution in Heroku was always to throw more money at it, and the increments were quite serious. With Hetzner, I now have massively overprovisioned servers that cost a fraction.
Because the server types you get for the price of a single Heroku dyno are incredibly beefy. And suddenly you need a lot less dynos. Which is quite important if you start managing them yourself.
There are internal reasons as well. Letting go of people can be highly disruptive and create uncertainty in your team. It’s a very unpleasant job that can also go wrong, especially if you have to fire loads.
Then there’s the perverse incentive that bigger teams usually equals a promotion. So if you’re the honest manager who manages a tight team and fires people, you won’t get promoted as often.
Top management knows this, of course. To sidestep these misaligned incentives a company-wide one-time layoff is really effective.
I absolutely love the idea of Zed, and I'm regularly giving it a go. Typing in Zed really feels better than VSCode. It's hard to describe, but impossible to discard once you've used it for a short while.
Unfortunately, there's a bunch of small things still holding me back. Proper file drag & drop for one, the ability to listen to audio files inside the editor, and even a bunch of extensions, in particular one that shows a spectrogram for an audio file.
Maybe my biggest gripe is that Python support is still better in VSCode. Clicking on definitions is faster and more reliable.
In vscode you can click on various assets, like images or audio files, and then view them right inside vscode. If you work with datasets, the ability to inspect them is crucial.
Yes ofc I can use Finder instead but in vscode I just cmd+p.
The reason it's faster is largely because it doesn't have all those little quality of life features and extension ecosystem. It's easyish to make software perform well if it doesn't do all that much. If you take base vscode, no extensions, and just do raw text editing, it's hard for me to tell the difference between vscode, zed, or any other editor.
When vscode was released, Sublime was faster - and it stayed faster. But that wasn't enough to stop the rise of vscode.
On desktop, you can just publish your software and slowly see it age as you work on your next big release. On iOS, it ages every year at brutal pace, and your new sales will plummet while you work on your next big release, meaning your revenue crashes much faster.
Even worse, the iOS App Store has no notion of paid upgrades, and publishing a new app is basically like starting from scratch as far as discoverability goes. So when you finally have your next big release ready, it's like launching a completely new company.
Apple really wants developers to make subscription apps that ship frequent iterative changes, and other business models just simply don't work well on their mobile platform (on Android it's even worse btw).