"lang" attribute not allowed in XHTML 1.1

[jmillikin@…]

In 0.5, the xml:lang attribute is copied to the lang attribute. However, in XHTML 1.1 this causes validation errors because lang is no longer allowed. The XHTML serializer should check which doctype is being serialized to before copying this attribute, or maybe even remove copying it from the XHTML serializer entirely (leave it in for HTML).

[cmlenz]

For XHTML 1.1 you should probably just be using the XML serializer, as you're giving up compatibility with HTML anyway. Or is there some feature provided by XHTML serialization in Genshi that is desirable for XHTML 1.1 output?

[jmillikin@…]

That's the solution I ended up using, but having the XHTML serializer write invalid XHTML is confusing enough that I figured it warrented a ticket.

[Carsten Klein <carsten.klein@…>]

Why not simply transform to xml:lang instead. It is supported by both XHTML1.0 and XHTML1.1.

[Carsten Klein <carsten.klein@…>]

Here is a proposed patch for the XHTMLSerializer that will transform existing lang attributes to xml:lang and will output xml:lang whenever there is a xml:lang attrib avail.

It also takes care of the rule that whenever xml:lang is available it takes over precedence above any existing lang attributes, see  http://www.w3.org/TR/xhtml1/#C_7

It also fixes the "bug" that when xml:lang was added to the buffer in the elif branch, it would be added a second time at the end of the if-elif statement.

output.py

@@ -347 +355 @@
                 for attr, value in attrib:
                     if attr in boolean_attrs:
                         value = attr
-                    elif attr == 'xml:lang' and 'lang' not in attrib:
-                        buf += [' lang="', escape(value), '"']
+                    elif attr == 'lang' and 'xml:lang' not in attrib:
+                        attr = 'xml:lang'
                     elif attr == 'xml:space':
                         continue
                     buf += [' ', attr, '="', escape(value), '"']

[Bjorn@…]

I have this problem, also in Genshi 0.6, except that I'm using  XHTML+RDFa 1.0 in the pages I generate. The lang attribute isn't allowed in XHTML+RDFa 1.0. There seems to be a plan to include it in  XHTML+RDFa 1.1, but I don't want to use that until it becomes a recommendation.

I tried the XML serializer, but then Explorer wouldn't display the generated pages. I have some empty script elements in my templates (pointing to separate Javascript files), and apparently Explorer doesn't understand that a script element is closed when it's represented as an empty tag instead of a start tag and an end tag, so it thinks the entire rest of the page is Javascript. Firefox and Seamonkey do the same if the pages are served as text/html, but display them correctly if served as application/xhtml+xml. With Explorer, application/xhtml+xml doesn't help. As much as I loathe being bug compatible with programs that won't follow standards, I do want even users of Explorer to be able to see my pages.

I think the right thing to do is to simply pass attributes through from the input to the output, and let users add both lang and xml:lang explicitly if they want both.

[Bjorn@…]

I propose this patch.

[anonymous]

XHTML1.1 was updated in 2010 to add the lang attribute back. I'm inclined to close this as "won't fix" unless someone presents a strong practical need to change Genshi's behaviour to comply with outdated versions of the XHTML1.1 specification.

See "Status of this document" in  http://www.w3.org/TR/2010/REC-xhtml11-20101123/ for a mention of this update to XHTML1.1.

[jmillikin@…]

Sounds reasonable. As long as Genshi's behavior matches the spec, it doesn't matter to me which one changed.