I think we bicker and pontificate about these kinds of issues too
much.
Do we want @ for every attribute or not? If we haven't decided,
we should decide.
If we do want @ for everything, then create a pull request which
does nothing but support @ for all current attributes, in
addition to the attributes without @. Don't worry about
"attribute sets" and other wild ideas for now. Just tackle that
one issue, otherwise no progress will be made. Attributes without
@ can become deprecation warnings, and can be removed later.
If you worry about the compiler becoming too complicated, I can
assure you it will barely have an impact on compilation speed,
and that's all users will care about.
If we don't want @ for everything, then propose something else.
Sorry to be curt, but we've talked about this for long enough and
gotten nowhere.
