> The protocol will be a simple sequence of messages, each message consisting of 32-bit size field in network byte order and the data. How hard it can be to implement such a simple specification? It turns out it can take months.
It could also take years, if not decades. It doesn't mean it should. This is a good article but this sort of straw-man shenanigans only serves to hurt its credibility.
It comes from experience. I've tried to implement such a simple protocol and it literally takes months to get all the details right. Also see the quote from Joe Armstrong in the article.
You aren't the only person who's well-versed in the network development and that statement is patently absurd. If it did in fact take you several months to get a framing layer right, the it only means that it was an issue stemming from specifics (and complexity) of your implementation approach. Hence my original credibility remark - you are clearly over-extrapolating and over-generalizing your anecdotal data.
It could also take years, if not decades. It doesn't mean it should. This is a good article but this sort of straw-man shenanigans only serves to hurt its credibility.