interoperability differences with RFC3986 section 2.3

I'm looking for feedback on this test case.  I imagine the browser 
behavior noted below has been well-known for many years, and I'm 
wondering - am I interpreting RFC3986 section 2.3 correctly when 
determining the pass/fail result of these tests.

"For consistency, percent-encoded octets in the ranges of ALPHA
    (%41-%5A and %61-%7A), DIGIT (%30-%39), hyphen (%2D), period (%2E),
    underscore (%5F), or tilde (%7E) should not be created by URI
    producers and, when found in a URI, should be decoded to their
    corresponding unreserved characters by URI normalizers."

Looking at "Test 1" hosted at:

http://lookout.net/test/uri/rfc3986-2.3.php

(FAIL) FF5         /%41%42%43/
(FAIL) Safari 5.1  /%41%42%43/
(PASS) IE 9        /ABC/
(PASS) Chrome 13   /ABC/
(FAIL) Opera 11.5  /%41%42%43/

My "Test 1" was to observe the resultant HTTP request generated from a 
simple reference to the following URI included as both an /href and an 
/img/@src

http://www.example.com/%41%42%43/

Best regards,
Chris Weber

Received on Wednesday, 10 August 2011 21:19:58 UTC