On Thu, 8 Feb 2018, Dennis Clarke wrote:
There is nothing wrong with RFC-3986 nor the more specific RFC-8089.
RFC 3986 is for generic URIs. RFC 8089 is for the specific subset file: URIs. They're different beasts.
The "wrong" about 3986 is that people and software are more and more often using URLs that violate that spec now.
The very fact that WHATWG is very browser focused causes me to ignore whatever they are doing.
I too am tempted to take that stand, but unfortunately I don't think that benefits our users much.
We occasionally see URLs being used on the web on the wild that "work in my browser" but they don't work in curl. They end up curl's problem either by users copying the URLs from the browser's address bar, users doing "copy link" or simply when asking curl to follow HTTP redirects - and more.
Over time we've (reluctantly) added adaptions when curl users have suffered. We now handle one, two or three slashes after the "scheme:" part, we url-encode illegal letters in redirect URLs (since people actually send such and the browsers deal with them) and so on. And I suspect we've not seen the end of those compromises.
URLs are not scoped to work within browsers *or* non-browsers. They work seamlessly across the entire Internet. They worked 20 years ago and I'm willing to bet they'll exist in another 20 years as well. The question is only exactly how to parse them... I think we as a community suffers as long as there isn't a one true URL spec.
I also work on a separate document where I try to nail down exactly what differences there are between the two - three primary URL specs:
https://github.com/bagder/docs/blob/master/URL-interop.md -- / daniel.haxx.se ------------------------------------------------------------------- Unsubscribe: https://cool.haxx.se/list/listinfo/curl-library Etiquette: https://curl.haxx.se/mail/etiquette.html
