[javanaut]

I don't know much about networks and haven't worked on any on-device software - I mostly work on element management systems.

Even on the high end there is a race to commodification. Router manufacturers have some similarity to server manufacturers like Dell - they get hardware and software components from 3rd parties and put them together. Your main bespoke software contribution might be device drivers and a data model.

High pay may not automatically translate into quality because there are other forces in play.

[nonameiguess]

There are at least a few use cases that can never be commoditized. My wife's ex used to do work for Wood's Hole developing firmware for acoustic routers intended to network submarines. Somewhat ironically, he thought it was just for scientific use but the US Navy was actually funding the development. This paid reasonably well for 15 years ago.

[javanaut]

Interesting! I don't disagree with you, but to jump off this...

I think there are some parallels to metro fibre networks. You have devices/pluggables with xxxG throughput. MUX/DEMUXs, ROADMs and ILAs are expensive. High cost, high margin.

But you don't make the optics. You're buying them from the same supplier as your competitors and you can't buy that company because you'll kill the market because your competitors won't buy from you and all you'd be left with is an interop problem. The second problem is that the market is small. Few outfits build these networks and they are often monopolies in their geo. There is little growth.

Commoditization is not the only problem.