W3C home > Mailing lists > Public > whatwg@whatwg.org > May 2009

[whatwg] External document subset support

From: Kristof Zelechovski <giecrilj@stegny.2a.pl>
Date: Mon, 18 May 2009 11:28:55 +0200
Message-ID: <EEA991D2BA954DDA81AC06EC827BAE79@POCZTOWIEC>
AFAIK, WebKit is not going to validate XML, they say it makes page load too
slow.  Besides, entities introduce a security risk because it can contain
incomplete syntax fragments and they can open a path to XML injection into,
say, <![DANGER[<span title="&malicious-entity;" >sweet kittens</span >]]>.
So XML processors often refuse to load cross-domain DTD or ENTITIES.

There are several XHTML entities that are indispensable for authors, namely
those that disambiguate characters are invisible or are indistinguishable
from others in a monospaced typeface.  These include spacing, dashes, quotes
and maybe text direction (deprecated).  Converting them to their
corresponding characters deteriorates the editing experience in an ordinary
text editor.  As far as codes for letters are concerned, text in a non-Latin
script would consist mainly of entities, which would make it extremely hard
to read, so this approach is not practical.  An editor limited to the ASCII
character set would be better off using a transliteration scheme and a
converter.

However, as some of the entities are indispensable, a DOCTYPE is required.
The browsers may support built-in entities but XML processors used to
process XHTML documents need not.  Providing a set of the entities needed
in-line is easy; however, the problem is that some validating processors
like MSXML require that the DTD, if provided, should fully describe the
document; providing entities only is not supported by default and the
processor refuses to load the document.  That means a DOCTYPE for XHTML is
necessary and should be provided by WHATWG (or by an independent party).
This DTD should be external in order to use parameter entities and, of
course, to make the document smaller.  It cannot, of course, define all
nuances of XHTML, but an upper approximation would be sufficient.  

The problem, of course, is maintenance, since XHTML is in flux.  XHTML is
currently described formally by a RELAX NG grammar and maintaining a
separate DTD would double the work to do so it would be best to be able to
generate the DTD automatically.  However, the converter I was advised to use
was unable to produce a DTD from the grammar because it is too complex for
the DTD formalism (of course).

Best regards,
Chris

Aside: Note that you cannot use DocBook with MSIE directly; a bug in the
default XSLT processor causes an error in initialization code.  This kills
all transformations, whatever your document is.  (I do not know about TEI.)
Received on Monday, 18 May 2009 02:28:55 UTC

This archive was generated by hypermail 2.3.1 : Monday, 13 April 2015 23:08:49 UTC