W3C home > Mailing lists > Public > xml-uri@w3.org > May 2000

PUBLIC v. SYSTEM not URI v. FPI

From: Rick JELLIFFE <ricko@geotempo.com>
Date: Thu, 25 May 2000 04:42:32 +0800
Message-ID: <392C3EB8.DF5C3F16@geotempo.com>
To: xml-uri@w3.org
 
> From: Liam Quin <liam@holoweb.net>

> Public identifiers are trying to solve a different problem than URLs --
> once a text has been given a public identifer, you always get the same
> text stream by dereferencing (in an unspecified way) a public identifier.
> 
> Public identifiers have the semantics we need for namespaces, and had
> namespaces been designed in the SGML world, there's no doubt in my
> mind that they would have been used for the purpose.
 
Actually, I think the important difference between a URL and an FPI is
not
anything in their syntax, it is the keywords PUBLIC or SYSTEM which
provide
a warrant about them.

As a syntax, they are little different:
   +//IDN www.sinica.edu.tw//DTD qaml.dtd//EN
versus
   http://www.sinica.edu.tw/dtd/qaml.dtd
in that they both have a scheme, a registered user, and then details.

FPIs may be nicer for the specific case of SGML/XML because they can
tell you "this is a DTD" or "this is a notation", but perhaps so can
a file extension in a URL, and most data in the world is NSGML (not
SGML)
anyway. So the syntax of URIs and FPI it is just "you say tomato and
I say tomato." 

I hadn't found the postings written by this "anti-URI" horde, so I 
cannot comment on them, but it seems that Tim BL or Dan C hold that 
using http: or urn: or whatever in a URL provides an equivalent warrant 
(about persistance, stability, uniqueness, etc) as the keywords PUBLIC
and
SYSTEM do in an entity or notation declaration in XML.   (Sorry if
I am putting words in their mouths here.)

I think they could, iff the RFCs for each access method defined
it as a property.  For example, if whatever RFC defines  ftp: 
said "The property of persistance should not be implied about
a URI that uses the ftp: method".  

I haven't checked all the RFCs, but I don't recall that they do
give this information.  So the problem is not URIs but that there
is no way to provide a warrant about them.  

<hobbyhorse>If my idea that
namespace URIs should be used as a base from which well-known
relative URLs could be used to locate various information, such
a warrant could be downloaded for a URL, using PICS I suppose.
Would that require too much of a descent from 50,000 feet? </>

Rick Jelliffe
Received on Wednesday, 24 May 2000 16:34:15 UTC

This archive was generated by hypermail 2.3.1 : Tuesday, 6 January 2015 20:32:42 UTC