441825 - XML won't load the style sheet if processing instruction contains an unencoded ampersand in FF3

Status: RESOLVED INVALID

(Core :: XML, defect)

Description

[Karl Arp]

User-Agent:       Mozilla/5.0 (Windows; U; Windows NT 5.1; en-US; rv:1.9) Gecko/2008052906 Firefox/3.0
Build Identifier: Mozilla/5.0 (Windows; U; Windows NT 5.1; en-US; rv:1.9) Gecko/2008052906 Firefox/3.0

I'm using CGI and PHP to deliver XSL stylesheets. When they contain an &, the following happens.
In Firefox 2, the XSL stylesheet loads just fine and the document is translated appropriately.
In Firefox 3, the stylesheet isn't requested, so it doesn't load, and Firefox displays the page as if each tag were just HTML.

In FF3, if you url encode the ampersand, it will request the stylesheet and all goes according to plan. However, according to w3c: "The ampersand character (&) and the left angle bracket (<) MUST NOT appear in their literal form, except when used as markup delimiters, or within a comment, a processing instruction, or a CDATA section. If they are needed elsewhere, they MUST be escaped using either numeric character references or the strings "&amp;" and "&lt;" respectively." (http://www.w3.org/TR/REC-xml/#syntax)

My understanding of that is that FF2 handles this correctly, while FF3 does not.

Reproducible: Always

Steps to Reproduce:
1.Open XML page with a link to a stylesheet which has an ampersand in the url.
2.
3.
Actual Results:  
Firefox 3 displays the content of the XML tags as if they were HTML tags.

Expected Results:  
Firefox 3 would translate the XML document using the XSL stylesheet and display the results.

I have several extensions installed, though I doubt they're the cause of the issue.
Firebug 1.2.0b3 *
Adblock Plus 0.7.5.5 *
ColorZilla 2.0
DOM Inspector 2.0.0
Gmail Manager 0.5.5
IE Tab 1.5.20080310 *
IE View 1.3.7
NoScript 1.6.9.3
Tamper Data 10.0.4
Web Developer 1.1.6
YSlow 0.9.5b2

This issue has been reproduced on another computer which only has the starred (*) extensions installed.

Comment 1

[Martijn Wargers (dead)]

Do you have a testcase somwhere were we can see the bug? Or perhaps, could you attach a testcase?

Comment 2

[Karl Arp]

Attachment: Here's a test case that fails in FF3, and should work in FF2

If this renders correctly, it will say:
Here is the first element
T1
Here is the second element
T2

If it fails to render correctly, it will say:
T1  T2


I obviously get the latter.

Comment 3

[Karl Arp]

I just thought I'd point out that, if you'd like to see the results as they should be displayed, you can just remove the arguments from the stylesheet link in the XML. i.e. "q.xsl?test=1&c=6" becomes "q.xsl"
It displays properly in both version of FF that way.

Comment 4

[Ria Klaassen (not reading all bugmail)]

Regression range is 
http://bonsai.mozilla.org/cvsquery.cgi?module=PhoenixTinderbox&date=explicit&mindate=2006-03-17+03%3A00&maxdate=2006-03-17+13%3A00
Seems likely that Bug 286132 is the cause.

Comment 5

[Jonas Sicking (:sicking) No longer reading bugmail consistently]

This seems to work according to spec. As pointed out in comment 0 even (or am I misunderstanding the comment?)

According to spec the '&' character in any pseduo-attribute in the xml-stylesheet processing instruction should be interpreted as the start of an entity. So:

<?xml-stylesheet href="foo.xml?a=b&c=d"?>

is invalid as 'c=d' isn't a valid entity, and it's also not terminated with a ';'.

<?xml-stylesheet href="foo.xml?a=b&amp;c=d"?>
is however valid and makes us fetch the url <foo.xml?a=b&c=d>

Similarly
<?xml-stylesheet href="foo.xml?a=b%26c=d"?>
is valid, but makes us fetch the url <foo.xml?a=b%26c=d> which is probably not what you want, since the server will decode that as a single query parameter "a" with the value "b&c=d".

Hope that makes sense.

Spec is available here:
http://www.w3.org/TR/xml-stylesheet/#NT-PseudoAtt

Comment 6

[Boris Zbarsky [:bzbarsky]]

Might be a good idea to add this testcase and variants to the regression suite, by the way, so someone doesn't come along and "fix" this.

Comment 7

[Karl Arp]

I just wanted to post this info in case anyone else was having the same problem understanding this as I did. In comment 0, I noted that the ampersand character must not appear in its literal form except when used as a processing instruction. This is why I thought it was something that needed to be fixed. After investigating some more to better understand why I was wrong, I found this link:
http://www.w3.org/TR/xml-stylesheet/#The%20xml-stylesheet%20processing%20instruction
It explains how the stylesheet processing instruction should be laid out, which, to me, seems to contradict the original statement. Either way, the second instance is specifically tailored to stylesheet PIs, and explains it fairly clearly.