fix(input[url]): Invalid URLS incorrectly valid

[btesser]

The current URL regex incorrectly matches invalid urls.
Examples:
http://www..
http:///
http://[[
The new regex should correctly identify urls

[btesser]

my fix is currently failing the test for:
http://server:123/asdfkj

Pretty much it's handling all cases except for domains without an extension. Any regex experts want to help?

[Narretz]

Url validation is currently following Chromium and FF, which is following rfc1035
In the latest snapshot, your examples are still valid. So I don't know if these changes will make it in or if they are even in rfc 1035. Maybe @caitp knows more.

[caitp]

I'm not sure we want a regexp like that in the tree, because that is kind of frightening. As I've said in another issue, URL validation in browsers is a bit more complicated and relies on quite a bit more than a simple regexp, so it's hard for us to be as robust as them.

I think there are some things we could add which would be worth doing, like supporting any scheme, or scheme-less urls (such as CDN urls), but there's a limit to how robust we can be when we're relying on a regular expression.

Even so, maybe someone else has a different opinion, @shahata WDYT about this?

[btesser]

at the very least I don't think the regex should allow things like http://www.. and http:///

The way I came across the bug was doing url validation for a form. Those are 2 probable mistakes that could be entered that I feel should be caught by the regex.

[Narretz]

There's been a change last year that made the url regex much less strict: e3be5d6 I haven't tested your examples, but I assume they are covered by this.