On 10/3/19 2:07 PM, Mark Andrews wrote: > Now IPv6 examples are nice but getting several 1000’s people to read draft > that > just add addresses in the range 2001:DB8::/32 instead of 11.0.0.0/8, > 12.0.0.0/8 > and 204.69.207.0/24, then to get the RFC editor to publish it is quite frankly > is a waste of time.
Long ago, I was working on Network Graphics Protocol and the draft RFC for it. My boss said that I needed to write a couple of paragraphs about fixed-point binary fractions, which the protocol used, "because that's not a common thing in the world". How bad was this? The person who was writing the generator of the NGP stream used *floating point* because that person didn't understand that 7FFFFFFF was interpreted as a bitty-point less than 1.0, and 80000001 was a bitty-point less that -1.0 -- and that trying to shoehorn this into IEEE floating-point format almost worked. What my poor boss didn't realize is that exactly 1.0 and -1.0 were not defined. The specification didn't make that clear. When I'm doing technical writing, I find all sorts of corner cases that were missed by the designers and the QA people. It makes me very unpopular. But it also makes for a better product, in the end. So making everything crystal clear and obvious is definitely not "a waste of time." You have no idea what undiscovered bugs may become obvious when you go through the exercise and show all your work. You still need a IPv6 version of RFC 1812. Make it as clean as possible. Use an ax instead of a XACTO knife on the current draft. What is the minimum necessary things that a generic IPv6 router MUST do?
