Re: XHTML character entity support

On Nov 4, 2009, at 10:56, Jirka Kosek wrote:

> As HTML5 spec doesn't define any DTD for its XML serialization you
> have
> to reference XHTML 1.x DTDs in order to use entities in it. It's
> implementation decision how to cope with external DTDs --

It's easier to implement compatible stuff if people don't go around
making their own decisions. :-)

> you can ignore them completely,

No, you can't if you want to read Web content, as WebKit has found.

> download them always with each request,

No, you can't if you want to have a performant app and want others on
your network to be able to access w3.org (by avoiding getting the IP
address of your firewall blocked due to excessive traffic).

> use XML catalog for redirecting to local copies of the most common
> DTDs,

If you do this, you get different attribute values than in browsers.

> use XML catalog for redirecting to only entity definitions instead
> of original full DTD,

So writing down how to do this would be helpful.

> I agree that having such possibilities listed somewhere together
> with a
> list of the most common system/public identifiers which should be
> recognized by UAs will improve interoperability.

OK.

> But such information is suitable for separate W3C Note. It has
> nothing to do with HTML5 spec.

(There's no such thing as W3C Notes. There are Working Group Notes.)

Why a Note and not a REC track publication?

>> As a practical matter, if I'm using SAX in Java, I can't get a
>> browser-style EntityResolver off-the-shelf as part of a common
>> org.apache package. (Or maybe I could but I'm unaware.)
>
> Is there anything you can't handle with XML catalogs:
>
> http://xml.apache.org/commons/components/resolver/

Having to research and write my own catalog file isn't materially
different from having to research and write my own .java file.

> I suppose that example XML catalog which maps known DTDs to just
> entity
> definitions would be nice addition to possible W3C Note I have
> mentioned
> above.

Implementors should be able to simply say |new
BrowserCompatibleEntityResolver()|.

-- 
Henri Sivonen
hsivonen@iki.fi
http://hsivonen.iki.fi/

Received on Wednesday, 4 November 2009 17:06:35 UTC