42278 - REC-html40 incorrectly returned as a result of reverse mapping XHTML identifier

Status: VERIFIED FIXED

(Core :: XML, defect, P3)

Description

[vidur (gone)]

Currently, we map both the REC-html40 and XHTML namespaces to the same internal 
identifier. The internal identifier is used in layout to special case handling 
of HTML elements. I believe we've decided to leave the forward mapping in place 
- we don't have to modify all the places that currently use the REC-html40 
namespace. Unfortunately, when we do the reverse mapping (map back from the 
common identifier to a namespace URI), we get the REC-html40 namespace. This 
means that this deprecated namespace will show up in operations such as 
serialization of XML documents. The reverse mapping should be fixed to return 
the XHTML namespace.

Comment 1

[vidur (gone)]

Nominating for beta2 so we can start the process of deprecating the old 
namespace. Also cc:ing Dave Hyatt to make sure that there's no dependency in XUL 
for the reverse mapping.

A patch is included below:

Index: nsNameSpaceManager.cpp
===================================================================
RCS file: /cvsroot/mozilla/layout/base/src/nsNameSpaceManager.cpp,v
retrieving revision 3.12
diff -c -r3.12 nsNameSpaceManager.cpp
*** nsNameSpaceManager.cpp      2000/04/03 11:11:08     3.12
--- nsNameSpaceManager.cpp      2000/06/12 23:14:02
***************
*** 110,133 ****

      nsString* xmlns = new nsString( NS_ConvertToString(kXMLNSNameSpaceURI) );
      nsString* xml = new nsString( NS_ConvertToString(kXMLNameSpaceURI) );
-     nsString* html = new nsString( NS_ConvertToString(kHTMLNameSpaceURI) );
      nsString* xhtml = new nsString( NS_ConvertToString(kXHTMLNameSpaceURI) );
      nsString* xlink = new nsString( NS_ConvertToString(kXLinkNameSpaceURI) );
      gURIArray->AppendElement(xmlns);  // ordering here needs to match IDs
      gURIArray->AppendElement(xml);
-     gURIArray->AppendElement(html);
      gURIArray->AppendElement(xhtml);
      gURIArray->AppendElement(xlink);
      NameSpaceURIKey xmlnsKey(xmlns);
      NameSpaceURIKey xmlKey(xml);
-     NameSpaceURIKey htmlKey(html);
      NameSpaceURIKey xhtmlKey(xhtml);
      NameSpaceURIKey xlinkKey(xlink);
      gURIToIDTable->Put(&xmlnsKey, (void*)kNameSpaceID_XMLNS);
      gURIToIDTable->Put(&xmlKey, (void*)kNameSpaceID_XML);
-     gURIToIDTable->Put(&htmlKey, (void*)kNameSpaceID_HTML);
      gURIToIDTable->Put(&xhtmlKey, (void*)kNameSpaceID_HTML);
      gURIToIDTable->Put(&xlinkKey, (void*)kNameSpaceID_XLink);
    }
    NS_ASSERTION(nsnull != gURIToIDTable, "no URI table");
    NS_ASSERTION(nsnull != gURIArray, "no URI array");
--- 110,133 ----

      nsString* xmlns = new nsString( NS_ConvertToString(kXMLNSNameSpaceURI) );
      nsString* xml = new nsString( NS_ConvertToString(kXMLNameSpaceURI) );
      nsString* xhtml = new nsString( NS_ConvertToString(kXHTMLNameSpaceURI) );
      nsString* xlink = new nsString( NS_ConvertToString(kXLinkNameSpaceURI) );
+     nsString* html = new nsString( NS_ConvertToString(kHTMLNameSpaceURI) );
      gURIArray->AppendElement(xmlns);  // ordering here needs to match IDs
      gURIArray->AppendElement(xml);
      gURIArray->AppendElement(xhtml);
      gURIArray->AppendElement(xlink);
+     gURIArray->AppendElement(html);
      NameSpaceURIKey xmlnsKey(xmlns);
      NameSpaceURIKey xmlKey(xml);
      NameSpaceURIKey xhtmlKey(xhtml);
      NameSpaceURIKey xlinkKey(xlink);
+     NameSpaceURIKey htmlKey(html);
      gURIToIDTable->Put(&xmlnsKey, (void*)kNameSpaceID_XMLNS);
      gURIToIDTable->Put(&xmlKey, (void*)kNameSpaceID_XML);
      gURIToIDTable->Put(&xhtmlKey, (void*)kNameSpaceID_HTML);
      gURIToIDTable->Put(&xlinkKey, (void*)kNameSpaceID_XLink);
+     gURIToIDTable->Put(&htmlKey, (void*)kNameSpaceID_HTML);
    }
    NS_ASSERTION(nsnull != gURIToIDTable, "no URI table");
    NS_ASSERTION(nsnull != gURIArray, "no URI array");

Index: nsIElementFactory.h
===================================================================
RCS file: /cvsroot/mozilla/layout/base/public/nsIElementFactory.h,v
retrieving revision 3.2
diff -c -r3.2 nsIElementFactory.h
*** nsIElementFactory.h 2000/05/10 13:13:05     3.2
--- nsIElementFactory.h 2000/06/12 23:15:38
***************
*** 46,52 ****
  #define NS_ELEMENT_FACTORY_PROGID          "component://netscape/layout/elemen
t-factory"
  #define NS_ELEMENT_FACTORY_PROGID_PREFIX   NS_ELEMENT_FACTORY_PROGID "?namespa
ce="

! #define NS_HTML_NAMESPACE                  "http://www.w3.org/TR/REC-html40"
  #define NS_XML_NAMESPACE                   "http://www.w3.org/XML/1998/namespa
ce"

  #define NS_HTML_ELEMENT_FACTORY_PROGID     NS_ELEMENT_FACTORY_PROGID_PREFIX NS
_HTML_NAMESPACE
--- 46,52 ----
  #define NS_ELEMENT_FACTORY_PROGID          "component://netscape/layout/elemen
t-factory"
  #define NS_ELEMENT_FACTORY_PROGID_PREFIX   NS_ELEMENT_FACTORY_PROGID "?namespa
ce="

! #define NS_HTML_NAMESPACE                  "http://www.w3.org/1999/xhtml"
  #define NS_XML_NAMESPACE                   "http://www.w3.org/XML/1998/namespa
ce"

  #define NS_HTML_ELEMENT_FACTORY_PROGID     NS_ELEMENT_FACTORY_PROGID_PREFIX NS
_HTML_NAMESPACE

Comment 2

[ekrock's old account (dead)]

This bug should be approved for nsbeta2+. Reason: developers will be 
using JavaScript to inspect XHTML elements (HTML elements in XML) to see what 
their namespace URI is (i.e. figure out which namespace a given element is part 
of), and we don't want them to be forced to hardcode recognition of the wrong, 
deprecated namespace REC-html40. If they do, then we'll be stuck with 
supporting that incorrect namespace mapping forever, and indeed it would be 
extremely difficult to ever fix the bug once content and web apps became 
dependent on the incorrect URI. Hence, fixing this bug is a critical requirement 
for assuring correctness of our XML Namespaces implementation.

Recommend nsbeta2+[6/22]. 

Comment 3

[ekrock's old account (dead)]

Note discussions of related issues in bug 7834, "REC-html40 should not be an 
HTML namespace".

Comment 4

[leger]

Putting on [nsbeta2+] radar. 

Comment 5

[David Baron :dbaron: (⌚️UTC-5, no longer working on Mozilla)]

I think rbs just fixed this by swapping the values of the constants, but that 
makes the constants misnamed.

Comment 6

[rbs]

I just noticed that Vidur has checked his patch in. Consequently, this bug is
going to be there because I had already temporarily swap the 'litteral strings'
of REC-html40 and the official namespace (bug 7834). I think it would be best
to remove the REC-html40 in nsNameSpaceManager.cpp and all the references to
kXHTML_... to avoid other potential collisions.

Comment 7

[vidur (gone)]

Ack! I didn't notice that rbs switched the values of the literal strings. I'd 
like to switch them back so that my patch works. I support the idea of removing 
the REC-html40 version and I know that rbs has gone through the effort of 
removing it from the Mozilla codebase. I suspect that we might break a host of 
demos that we've already created using the old namespace, but now's the time to 
update them, right? Eric Krock, what do you think?

I'm going to switch back the values of the literal strings and wait for ekrock's 
response before removing the REC-html40 namespace.

Comment 8

[vidur (gone)]

Switched back the values of the string literals. I'm going to close this one and 
leave the discussion on whether to get rid of the REC-html40 namespace in bug 
7834.

Comment 9

[Chris Petersen]

Marking verified based on last comments.