Domains: Improve copy for TLDs unavailable for registration on WordPress.com

[klimeryk]

Fixes #6810

Before, the error message was not very helpful (and it was shown as an intimidating error):

Now, the copy has been improved (thank you, @ranh!) and the notice has info severity:

The Learn more link points to https://en.support.wordpress.com/map-existing-domain/.

/cc @aidvu @umurkontaci for code review
/cc @spncrb

Test live: https://calypso.live/?branch=fix/tld-error-message

[ghost]

Product looks good to me 👍

[umurkontaci]

What about SLDs? (.co.uk e.g.)

[ghost]

that's a good point since .co.uk results in

[umurkontaci]

There's a isSubdomain function in https://github.com/Automattic/wp-calypso/blob/master/client/lib/domains/index.js#L114 which might be useful

[klimeryk]

Yes, I realized that it will do that for SLDs - but the sentence remains true. The domains "ends with .uk". I'm not familiar with any SLD which we don't allow registrations but allow registrations for the TLD of that SLD (ugh, hope that's clear). So, for co.uk we don't support .co.uk nor do we support .uk.
Why not just use the SLD in the notice? I can't figure out an easy way to reliably extract SLDs, if present. Picking just the TLD is safer.

[umurkontaci]

The whole point of the message is about the domains we don't support though. A straightforward, non-optimized way could be to count the dots and make use of the isSubdomain function;

if (dots.length == 1) {
  // TLD
} else if ( isSubdomain( domain ) ) {
  // TLD (also subdomain)
} else {
  // SLD (grab last 2 dots)
}

[klimeryk]

The reason why I've not opted for such a solution is because unfortunately it's not as straightforward (as it usually is with domains). The isSubdomain function is very simple and dumb - it checks for 2-3 characters long tokens between dots and assumes those are TLD/SLDs. But there are many >3 character long SLDs (.school.za, .csiro.au, .film.hu, etc.) and there are many 2-3 character long domain names (cnn.com is an obvious one, az.pl is another example). The only reliable way is to have a hardcoded list of all the possible SLDs - something we can't afford on the frontend and doesn't seem worth a trip to the backend. Hence my proposal to just use the TLD part. It's not ideal, but I think given the situation is the optimal solution (it's better to stick to something we are more certain of than to confuse the user with guessing if we're dealing with a TLD/SLD/subdomain).

[umurkontaci]

I still think handling co.uk. is more important than some super rare domains. But if you are feeling strongly towards handling it this way I think that's alright as well.

[aidvu]

You could add this info in the error response from the server? You get the TLD extracted there anyway.

[klimeryk]

Yeah, that's probably the best to do it - I'll try to do that in a separate patch in the meantime (though, I admit, it's not high priority 😉).

[umurkontaci]

Our node version has been bumped to 6.1.0 since this branch has been created. So it needs a rebase :)

[umurkontaci]

I did the rebase everything seems OK.

This is good to go, is there possibility of a delay when we try to determine whether the domain is registered or not? Like when the user registers a domain elsewhere and comes to our site right after and sees this message? In a sense, they already came here to "add" their domain.

It's certainly better than what we had before but I think it's something to think about for future.

[klimeryk]

That's a good point. It's probably highly dependent on the registrar - how fast they update/propagate. @spncrb, can you keep an eye out for such cases and let me know if they are confusing for users?