On 11/17/2020 9:56 AM, Ian Swett wrote: > I just want to say that in Software Engineering, specification > teams and implementation teams are usually different, that is my > background. > As I said this is a deep issue. Certainly it is good the chairs > pay attention to it. > > > I agree that's all too often the case, but I don't see how one can be > sure a spec works if the specification team is separate from the > implementation team?
That has been an ongoing issue in the software industry for a long time. The old 1980-IBM-style "cascade" model supposed a team of analysts studying the market and the customers requirements, and then a somewhat subservient team of programmers studying the spec and producing the code. To complete the cascade model, the less regarded team of testers would verify that the code actually implements the spec. Let's say that the successful part of the software industry has moved away from that model, and uses now constant feedback models in which specs are updated with feedback from development, tests and user trials, and in which there is definitely not a hierarchical relation between program managers, developers, and quality insurance. This is precisely what the QUIC WG did: there was lots of feedback throughout the process, based on experience with networks, deployment on small systems and in server farms, interop testing, etc. I wish more WG would use such processes. -- Christian Huitema
