From: Daniel Martin <martin@...>
Date: 2007-05-19T03:16:25+09:00
Subject: Re: URI::merge bug?

Nobuyoshi Nakada <nobu@ruby-lang.org> writes:

> Hi,
>
> At Sat, 19 May 2007 00:52:37 +0900,
> Daniel Martin wrote in [ruby-talk:252116]:
>> You know what?  I'm using an old RFC.  Everything I said applies to
>> RFC 2396, but that's not the current URI RFC.  The current one is
>> RFC 3986, and by that one firefox is doing exactly the right thing.
>> (So, by the way, is Opera)
>> 
>> So it is a bug in URI, and the bug is "written to an old RFC".
>
> Then, the test is wrong too?

Yes, it is.  That test case is straight out of RFC 2396, and it is
clearly incorrect given the pseudocode in section 5.2.2 (pages 31-32)
of RFC 3986.

RFC 3986 updates the test cases, including them in section 5.4  (Not
in an appendix as RFC 2396 did)  That test case is mentioned, and the
behavior is indeed what it looks like from the pseudocode.

We should probably check that ruby's uri module passes all the tests
now in RFC 3986.  If I have a chance I'll go do that.

-- 
s=%q(  Daniel Martin -- martin@snowplow.org
       puts "s=%q(#{s})",s.to_a.last       )
       puts "s=%q(#{s})",s.to_a.last