Alan de Freitas

We have something in the quick look about `pct_encoded_view`. Should we interleave this content with "parsing"? Or a section before containers?

The term `target` is kind of problematic because it can be more than `origin_form`. ``` request-target = origin-form / absolute-form / authority-form / asterisk-form ``` and ``` urls::result< urls::url_view >...

So we ask Peter?

It sounds inconvenient when we know or don't care if the thing exists: ```cpp auto h = u.host(); ``` becomes ```cpp auto hr = u.host(); pct_encoded_view h; if (hr) h...

Not always. It's always an empty placeholder when `!u.has_authority()`. Anyway, we also have `has_scheme()`, etc... which could be just as inconvenient.

The difference is when an empty string is an acceptable replacement for a missing component in that application. Especially if the user knows where the URL is coming from. ```cpp...

> ! u.scheme().empty() is just as good as u.has_scheme() `! u.scheme().empty()` implies it could be empty. > mentioning in the javadoc that if there is no scheme, the return value...

I've been thinking about this. In particular why I thought it was a good idea to use optional in the magnet link example. And I guess the reason is that...

> library consumes as little as possible So the criteria are - Included headers - Binary size - Size of stack buffers - Dynamic allocations Any other criteria? About the...

OK. So the issue converges to - Macro to disable source locations in error codes and exceptions - Macro to tune the size of internal stack buffers