xml-dev - Re: URNs as SYSTEM IDs

["Simon St.Laurent" <simonstl@simonstl.com>]

At 04:26 PM 11/23/00 -0500, John Cowan wrote:
>On Wed, 22 Nov 2000, Simon St.Laurent wrote:
>> _If_ the DTD makes no essential Infoset contributions, the DTD doesn't
>> matter at all except to provide constraints.  However, since DTDs
>> frequently make Infoset contributions (and the IANA registration document
>> sets no such limitations), processors have no way of knowing whether or not
>> the DTD is required for proper processing.
>
>That is the purpose of the standalone declaration.

Perhaps, but the standalone value does absolutely nothing unless it is set
to yes and an external reference is made anyway.  standalone="no" does
absolutely nothing, per the XML 1.0 spec.

If the spec so much as required parsers to report errors when standalone
was set to no and the parser skipped or was unable to retrieve external
resources, I might have more sympathy for your position. It does not - it
only requires reporting when a 'yes' declaration is apparently violated.

>Let me drive the stake straight through the vampire's heart:
>
>	The use of URLs as system identifiers in XML means
>	that the retrieval of remote parts of an XML
>	document can *never* be reliable.
>
>URLs are part of the Web, and the essence of the Web is that it is
>a best-effort resource delivery mechanism.  It is not, and never
>can be, reliable: servers go down, choke routers fail, documents
>are moved by well-meaning but blundering administrators, and so on.
>
>To design reliable systems with the Web at their core is to build
>ropes out of sand.

There is nothing absolutely reliable on the Web.  There are, however,
degrees of reliability.  URNs are not presently reliable at all when used
in open Internet systems.

It's a stake, yes, but not one that actually kills the vampire.

Sorry!

Simon St.Laurent
XML Elements of Style / XML: A Primer, 2nd Ed.
XHTML: Migrating Toward XML
http://www.simonstl.com - XML essays and books