I think there's something like a dual of the uncanny valley when it comes to protocol complexity vs adoption. Really simple like UART, I2C, or SPI, and engineer

I think there's something like a dual of the uncanny valley when it comes to protocol complexity vs adoption. Really simple like UART, I2C, or SPI, and engineers will adopt it on their own. But once you start wanting to add some higher level features, engineers would just as soon reinvent their own proprietary schemes for their own specific needs (the "valley"), and the network effects go away. So to create a more popular protocol you end up with a design committee where everyone piles in their own bespoke feature requests and the thing ends up being an ugly complex standard that nobody really likes. But at least it gives it a shot at wider adoption to prime the pump of network effects. (Or maybe it's more akin to how Java beats the Lisp Curse by using the social effect of having to raise an army to do anything?)