Want a Rant? (RTOS and "libraries" vs. the dark side)

[peter-h]

In the case I'm thinking of it is literally one guy, an external contractor, and the thing they're asking him to work on has nothing to do with their products.  Imagine something like a network stack on a refrigerator (not the actual product but a placeholder) from a refrigerator manufacturer, any time above zero spent on mucking around with the network stack internals is wasted effort because it'll never be touched again once it's up and running, their business is refrigerators, not networking software.

I don't think you can do that indefinitely into the future.

You can make the same refrigerator "for ever" but external connectivity stuff will need a re-visit. The original dev will be long gone...

I spoke to a customer today who bought a load of my product (a user-programmable box) 15-20 years ago. The dev (who wrote the code for him) is so long gone they don't even know who it was.

I don't think any of this has a good solution, but if they did it in-house they would have a better chance, even if the approach was sub-optimal. But maybe not; most companies lose everything after 5-10 years

[westfw]

one guy, an external contractor, and the thing they're asking him to work on has nothing to do with their products.  Imagine something like a network stack on a refrigerator

This is not dissimilar to how a lot of consumer are created.  Company "USAcomp" says to some contractor "Cont2ndW"  in some country where development labor is cheap and says "we want a box that does X; it's just this generic SoC reference design running linux, with this special ZChip, but we want the user interface to look like Y."  Product quality then depends on the choice of Cont2ndW, how close the relationship is, and how proaactive USAComp is about getting bugs fixed.