I noticed a lot of threads about how it's good that this (apparently more reliable than K-9 or gmail) app needs to be thrown out and rewritten because of android-side API drift.
There's an effect called "survivor bias", which for software, usually means that the older a popular application, the better it is. For instance, Unix shell has survived more or less intact for 50 years. That implies no significantly better CLI language has been invented by the last two generations of programmers.
This idea of just deleting quality software every five years for no reason is setting the field back.
Imagine what things would look like today if Microsoft had managed to replace 100% of unix shell implementations with the DOS command prompt!
Indeed - what a lot of comments on this topic seem to overlook is that a lot of device specific bug fixes have been put into effect for push sync in FairEmail, without battery drain, and while only using IMAP IDLE (therefore avoiding cloud sync or sending email passwords to third party services for push messages).
The idea of rewriting that every few years in a new language, or using new APIs, is counter to the principle of writing robust, reliable software without regressions. In my experience, those writing software that reinvents the wheel every 5 years end up ignoring and externalising the impact of regressions on their users, for the sake of using a new language or framework every few weeks...
"The new shell is more advanced, so it's better!" is false equivalence bias. It's more complex, which isn't the same thing as better.
Shells are user interfaces, not programming languages. This user interface has been good enough that we have not yet needed to invent a small, portable, ubiquitous, systems programming language [sans user interface]. The interest in making a new more complex shell, rather than a programming language, is the survivor bias. Because we've used user interfaces for systems programming for eons, and that use case hasn't died yet, we must just need a more advanced shell, rather than a different solution.
> Unix shell has survived more or less intact for 50 years. That implies no significantly better CLI language has been invented
False. Network effect is an equally good explanation. In fact, one of the arguments that repeatedly crops up on HN in favour of bash is "it's installed already".
The first thing I install on a Linux box is powershell. Your mention of the DOS shell is a straw man - it hasn't been state of the art on Windows for nearly twenty years.
So you picked a bad example but your point has merit? No. I don't think it does. Kernel devs see problems with fork. Fork is thirty years old, maybe forty? Sometimes things need to change.
There's an effect called "survivor bias", which for software, usually means that the older a popular application, the better it is. For instance, Unix shell has survived more or less intact for 50 years. That implies no significantly better CLI language has been invented by the last two generations of programmers.
This idea of just deleting quality software every five years for no reason is setting the field back.
Imagine what things would look like today if Microsoft had managed to replace 100% of unix shell implementations with the DOS command prompt!