W3C home > Mailing lists > Public > www-html-editor@w3.org > January to March 1999

Re: XHTML Working Draft--URLs for system ids of char entity decls

From: Paul Grosso <pgrosso@arbortext.com>
Date: Mon, 22 Mar 1999 18:28:10 -0600
Message-Id: <3.0.32.19990322182759.00d19ff4@pophost.arbortext.com>
To: Murray Altheim <Murray.Altheim@Eng.Sun.COM>, www-html-editor@w3.org, w3c-html-wg@w3.org
Murray,

Thanks for the reply.  I don't want to rehash the WG's discussions,
but I can't help but make some comments embedded below.

At 15:52 1999 03 22 -0800, Murray Altheim wrote:

>This was discussed at some length, and the decision to not use absolute
>URIs was made for two reasons:
>
>   1. Because the XML Recommendation does not specify whether the 
>      publicId or systemId has precedence, there was concern that
>      if XHTML authors were to simply use the URL as provided (which
>      I believe the vast majority would do), there may be an 
>      unnecessary and draining load on the W3C http server each 
>      time an XHTML document is validated. This may be considered
>      a more fragile situation than providing a relative URL.

I don't think these are good reasons.  The XML spec doesn't specify
a precedence precisely because the processor should be allowed to do
what makes sense for it.  Assuming, then, that a processor handed both
a public id and a system id would necessarily do something stupid is
not a good reason for anything.

>      
>   2. Providing an absolute URI might indicate that documents not
>      using the provided one were somehow not conformant, 
>      unintentionally canonicalizing the URI itself in the spec.

I don't see this.  I don't think providing an absolute URI says
any such thing, but if you do think someone might misinterpret
it to be so saying, then put a note into the spec disabusing them
of their misunderstanding.

What could possibly be the role of the public id if it were an
error to do anything but use the system id?

>
>I believe it better to advocate that authors and others keep a 
>copy of the DTD in the same directory as the document, and if
>they wish to put it in a different directory in their environment,
>that they modify their DOCTYPE's URIs to point to that new location.
>Or use a catalog file (which is what I'd recommend if XML had them
>in the spec).

I believe it is both bad practice and folly to assume that the DTD
and all its referenced entities will follow the document around the web.

I agree that the use of a catalog is a good idea--in fact, that is what
started my looking into this.  I use the catalog to map the public id
into my local copy, but I want to know that the system id will be
resolvable when I send out a document to someone else on the web.
Am I confused, or isn't this the whole point of XML on the web?

>I don't think this is a great solution, but it seems that the 
>alternative might be worse, especially if the W3C server was hit 
>ceaselessly by each opening XHTML document.

Why should a URL to the DTD's entities cause any more hits than 
the URL to the W3C graphic that appears all over the place?  Don't
tools out there do caching?  Or, if the tools want to "hardcode"
knowledge of these entity references into themselves, that's fine
too.  But why make masses of individual, potentially naive, users 
explicitly download a bunch of entities files to their local machine 
and have to put them in a particular directory just to be able to 
open a document in their browser?  You can't be suggesting this.  
I must be missing something.

paul
Received on Monday, 22 March 1999 19:28:49 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Tuesday, 27 March 2012 18:16:44 GMT