Hi Rowan
>
> Rather than a *representation* of a URL, think of the class as a
> *builder* for URLs. There are multiple methods because you might want to
> build the URL in different orders ("start with this URL but replace the
> port", "start with this domain and I'll add the path later", etc). All
> the flags are related to how the input should be interpreted, and the
> output manipulated, in order to build a correctly formatted URL string.
>
Sorry if it was really not clear in the RFC since I didn't even talk about
the CURLOPT_CURLU option, but this class is not only there to parse/build
strings for Curl but to give this specific object to Curl instead of a
string representation of an URL.
>
> Maybe it should even be called CurlUrlBuilder? That also fits with the
> design of having mutable setters; as Derick pointed out, mutable value
> objects are generally a bad idea, so it would make sense to encourage
> users to think of this as a way to get one or more strings, rather than
> as a result in itself.
>
Since you can give this object to curl instead of an URL string, I would
not call it CurlUrlBuilder.
Regards,
Pierrick
