[RFC] type safe trees

[Richard Kenner]

    Is that because you think that all of the C++ language would be
    genuinely useful in writing the compiler, or because you think
    that people would use language features that aren't useful?

    Again, I find this argument a bit peculiar.  It relies on thinking
    that some language features would be demonstrably harmful,
    but also that there would be no reason for anyone to avoid
    using any language features.  I don't see how those arguments
    can both be true.

Because "harmful" and "useful" aren't binary.  And because the people who
find it "useful" and those that find it "harmful" are different.

There will always be a tradeoff with each proposed expansion of the subset.