GURL: add a HostNoBrackets() method for bracketless IPv6 literals.

A URL can contain an IPv6 address literal, like [2001:db8::2].  Users of
the GURL library sometimes need the square brackets, and sometimes don't.

My proposed patch adds a HostNoBrackets() method, which returns the hostname
with brackets chopped off.

Tested using the Chromium source tree:
hammer -C build googleurl_unittests --mode=Release
hammer -C build googleurl_unittests --mode=Debug

[brettw, 2009-05-10 14:50:53]

When would we need this function as opposed to the regular "host" function? It
would be nice to find a name that better indicates when people should use it:
PlaintHose seems pretty generic to me.

[pmarks, 2009-05-10 20:20:48]

On 2009/05/10 14:50:53, brettw wrote:
> When would we need this function as opposed to the regular "host" function? It
> would be nice to find a name that better indicates when people should use it:
> PlainHost seems pretty generic to me.

The use case I have in mind is for sending to HostResolver/getaddrinfo() to get
the address in binary form.

See this codereview, where wtc suggested that the bracket-stripping be done in
the GURL library:
http://codereview.chromium.org/113026/diff/1/3

See this issue for some discussion about the name:
http://code.google.com/p/google-url/issues/detail?id=2

HostForRequest() was initially suggested, but I realized that the "Host:"
parameter in an HTTP request should have the square brackets included, so that
would be a deceptive name.

My opinion is that this function is sufficiently generic that we should name it
based on what it does, not on how it's used.  There may be other uses for a
bracket-less host that I haven't considered yet.

Feel free to pick an alternate name.  Here's some of my ideas:
- PlainHost
- RawHost
- BracketlessHost
- HostNoBrackets
- NakedHost
- InnerHost
- StrippedHost
- LiteralHost
- HostForAddress

[eroman, 2009-05-11 18:27:54]

As an aside, it might be useful if the canonicalization step would save an enum
for what type "host" was (PLAIN, IPV4, IPV6), then this function could look to
that enum rather than scan for brackets. (similarly the IsIpAddress() wouldn't
need to re-canonicalize)...

[eroman, 2009-05-11 18:37:47]

sounds good to me.

[sparkmaul, 2009-05-11 18:53:30]

I've renamed PlainHost -> HostNoBrackets.

As for the enum change, I think we should define the interface and get a simple
implementation first, and defer wider optimizations to a future CL.

[brettw, 2009-05-11 19:16:24]

LGTM. I don't see a burning need for the enum optimization unless we have more
use cases for it in the future.

[pmarks, 2009-05-12 22:27:57]

One of you will need to commit this for me.  I don't have commit access to
chromium or googleurl.