Why isn't dip1000 fully implemented yet?

[Daniel N]

On Thursday, 22 November 2018 at 10:12:12 UTC, Nicholas Wilson 
wrote:
>
> dlang.org#2453 is a blocker for that, and _not_ the other way 
> round, for the sole reason that without knowing what it is 
> supposed to do, it is impossible to review (the bugzilla issues 
> are not helpful).
>
> Note that dlang.org#2453 also documents other changes to 
> dip1000 that walter has made and has chosen not to document 
> (just recently I had to inform Razvan who is otherwise a very 
> capable compiler developer what the situation was because there 
> is no documentation), which makes the whole process even worse 
> because the implementation bears less and less resemblance to 
> the documentation and keeping track of interactions gets 
> combinatorially more difficult, and when reviewers are faced 
> with that they simply just do not review.

"impossible to review" is a very strong statement, how then were 
you able to write the spec? I had no issues understanding what 
Walter meant both comments and the test cases speak for 
themselves, he also offered to answer any questions.

If there is any corner case which you don't understand, it would 
imho be more constructive to ask him to add a new test case for 
it in #8504 so that any ambiguity will be cleared, future 
regressions prevented and coverage improved. It's very easy to 
misunderstand a written explanation, but code can only be 
interpreted one way. This is even more so for an international 
project where not everyone is a native English speaker.

In my world the spec/doc/changelog is just a release blocker for 
the next official external compiler release, but by all means 
merge them simultaneously, now that they both are available.