Sun's URL class has bug in creation of relative query URLs
----------------------------------------------------------

                 Key: NUTCH-566
                 URL: https://issues.apache.org/jira/browse/NUTCH-566
             Project: Nutch
          Issue Type: Bug
          Components: fetcher
    Affects Versions: 0.9.0, 0.8.1, 0.8
         Environment: MacOS X and Linux (CentOS 4.5) both
            Reporter: Doug Cook
            Priority: Minor


I'm using 0.81, but this will affect all other versions as well.

Relative links of the form "?blah" are resolved incorrectly. For example, with 
a base URL of http://www.fleurie.org/entreprise.asp, and a relative link of 
"?id_entrep=111", Nutch will resolve this pair to the link
"http://www.fleurie.org/?id_entrep=111";. No such URL exists, and all browsers I 
tried will resolve the pair to 
"http://www.fleurie.org/entreprise.asp?id_entrep=111";.

I tracked this down to what could be called a bug in Sun's URL class. According 
to Sun's spec, they parse the relative URL according to RFC 2396. But the 
original RFC for relative links was RFC 1808, and the two RFCs differ in how 
they handle relative links beginning with "?". Most browsers (Netscape/Mozilla, 
IE, Safari) implemented RFC 1808, and stuck with it (for compatibility and also 
because the behavior makes more sense). Apparently even the people that wrote 
RFC 2396 recognized that this was a mistake, and the specified behavior was 
changed in RFC 3986 to match what browsers do. 

For a discussion of this, see  
http://gbiv.com/protocols/uri/rev-2002/issues.html#003-relative-query

Sun's URL implementation, however, still implements RFC2396, as far as I can 
tell, and is out of step with the rest of the world.
This breaks link extraction on a number of sites.

I implemented a simple workaround, which I'm attaching. It is a static method 
to create URLs which behaves exactly as new URL(URL base, String relativePath), 
and I use it as a drop-in replacement for that in DOMContentUtils, Javascript 
link extraction, etc. Obviously, it really only matters wherever links are 
extracted. I haven't included the calling code from DOMContentUtils, etc. 
because my local versions are largely rewritten, but it should be pretty 
obvious.

I put it in the org.apache.nutch.net directory, but obviously feel free to move 
it to another place if you feel it belongs there!




-- 
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.

Reply via email to