Especially when the topic of said call is “uh, why doesn’t your chip actually, you know, work?”. Let’s call the vendor ABC pretend for the purposes of this post that the hardware in question is a network switch.
-
ABC Engineer: Oh, your network uses ABC 123 NICs? You should probably upgrade to the ABC 456 NICs.
Us: Look, that isn’t the problem. We can reproduce the issue with any NIC.
ABC: Oh, yeah, that’s not causing this problem, but you really should think about using 456s.
2)
ABC: Oh, you’re using our switch to actually make a network? Most of our customers use the switch as a repeater between two computers.
Us: So the majority of your testing has gone into making sure that your switch can properly act as a wire?
3)
Us: So, just how many of these switches have you actually sold, anyway? Lots, right? How are we the first people to see this?
ABC Sales Guy: We’ve sold thousands – no, TENS of thousands of these things.
ABC Engineer: Hey, look, you really should think about moving to the 456…
Bonus Comment from one of our hardware engineers(not the vendor), trying to convince us to re-implement all of our software to be more fault-tolerant in the face of completely broken hardware:
Engineer(completely serious): It’s unrealistic to expect the hardware to actually work.