Always return normalized urls in lower case

[walro]

This would avoid problems with Autoping as it currently requires at least the protocol to be in lower-case.

I can't think of any drawback of doing this, can anyone else?

[dentarg]

From http://stackoverflow.com/questions/7996919/should-url-be-case-sensitive

Depends on the hosting os. Sites that are hosted on Windows tend to be case insensitive as the underlying file system is case insensitive. Sites hosted on Unix type systems tend to be case sensitive as their underlying file systems are typically case sensitive. The host name part of the URL is always case insensitive, it's the rest of the path that varies.

Another answer cites W3:

There may be URLs, or parts of URLs, where case doesn't matter, but identifying these may not be easy. Users should always consider that URLs are case-sensitive.

[walro]

Yeah, true. How about just the protocol part? I'll change the issue title in that case.

[jage]

• Scheme name: normalize to lower case
• The rest: Don't touch

[walro]

Sounds good to me

[dentarg]

http://stackoverflow.com/questions/2148603/is-the-protocol-name-in-urls-case-sensitive

The scheme is case-insensitive: http://tools.ietf.org/html/rfc3986#section-3.1

The hostname is also case-insensitive, since it's dns.

The rest is case sensitive.

[jage]

Actually, I've changed my mind.

Lower case everything, but don't use it for other than comparison. Always save the original URL (with lower case protocol), and use that when requesting or linking to the resource.

[dentarg]

@jage Uhm, this is a library, it doesn't save anything?

[jage]

@jage Uhm, this is a library, it doesn't save anything?

Yes, I'm thinking of when you're using this library. We need to separate normalized URLs and original URLs.

[jage]

So, lower case scheme and hostname. Write a test and code that makes the test pass, then we can close this issue.

[dentarg]

I'm thinking of when you're using this library.

Isn't that a discussion we should have in another issue? Or do you propose additional functions to this library?

[jage]

Isn't that a discussion we should have in another issue?

No need for a special issue.

Since we're discussing it here I thought it would be suitable to give my thoughts about how to actually use the normalized URLs.

Or do you propose additional functions to this library?

No.

You can't reason about how this library should act unless you talk about how it should be used.

[dentarg]

Closing this in favor of #9, because I think that is more clear than changing the title of this issue. I like our issues short and precise.

If we want the hostname in lower case, I think we should open a new issue about that. I'm not sure we should add that functionality just because.

It's a good idea, that @jage mentioned, to be able to do comparisons, but I think we can open a new issue on this repo to discuss that.

[dentarg]

Lower case everything, but don't use it for other than comparison. Always save the original URL (with lower case protocol), and use that when requesting or linking to the resource.

I agree with this. I think it's the correct behavior for this gem.

I think I had a hard time agreeing before, because we are doing it wrong in zambezi, we don't save original the URL, we save the normalized URL.