> Oh, it looks like you are right: useless strings are already removed 
> during compilation. But it looks a little bit surprising to me to use a 
> multiline string for a comment. I prefer classic # comments.

I was surprised by this as well.  I think the comment in _header_value_parser was originally a docstring and got moved around during editing.  It should probably get changed to a block comment.

I could have sworn that when I learned about the optimization it was in the context of documentation that specifically said the optimization was done so that triple quoted strings could be used as multi-line comments, but I cannot find it through google or guessing where I might have seen it in our docs, so perhaps I was imagining things.