Use NCR instead of HTML entity in feed autodiscovery links

[peaceablewhale]

Currently, » is used as the separator. This will break sites served with the internet media type "application/xhtml+xml". The equivalent numeric character reference » should be used instead.

[westi]

Why does this break the page when serving as XHTML?

AFAIK XHTML supports the same named entities as HTML4 which would inclide &raquo'.

​http://www.w3.org/TR/html401/sgml/entities.html

Changing this makes it less clear what it being done.

[nacin]

application/xhtml+xml != XHTML. See also #14225, #14226.

[westi]

In general we don't code for serving as application/xhtml+xml but rather as text/html

As far as entities are concerned the named entities are fine in XML as long as the parser used can cope with them:

​http://en.wikipedia.org/wiki/List_of_XML_and_HTML_character_entity_references#Entities_representing_special_characters_in_XHTML

[peaceablewhale]

@westi:
The XHTML 1.x specification did adopt all entities from HTML 4 and place them in the DTDs. However, browsers and most XML processors currently do not load DTD (at least by default) and therefore the entity references cannot be loaded properly and fatal error occurs.

In addition, HTML 5 does not provide a DTD nor require browsers to load a DTD. The ​HTML/XHTML Compatibility Authoring Guidelines stated that no entity reference is used.

@nacin:
A XHTML document should be served as application/xhtml+xml. By doing so, SVG and MathML can be embedded to the XHTML document inline. Internet Explorer 9 has support to application/xhtml+xml.

[peaceablewhale]

If using NRC directly makes the code less clear, ent2ncr() may be used for the conversion.

[nacin]

Do we still care about this? Please re-open if so.