W3C home > Mailing lists > Public > xsl-editors@w3.org > July to September 2004

FW: 1.1 WD Comment on autogenerated property IDs.

From: Paul Grosso <pgrosso@arbortext.com>
Date: Wed, 29 Sep 2004 09:40:17 -0400
Message-ID: <F13E1BF26B19BA40AF3C0DE7D4DA0C03A5C5E4@ati-mail01.arbortext.local>
To: <xsl-editors@w3.org>
Forwarding to xsl-editors.

________________________________

From: w3c-xsl-fo-sg-request@w3.org [mailto:w3c-xsl-fo-sg-request@w3.org]
On Behalf Of Steve Zilles
Sent: Tuesday, 2004 September 28 20:53
To: Glen Mazza
Cc: w3c-xsl-fo-sg@w3.org
Subject: 1.1 WD Comment on autogenerated property IDs.


On Sat, 21 Aug 2004 11:14:02 -0400, Glenn Mazza wrote:



	Editors:
	
	I previously wrote [1] on the issue of autogenerated property
ID's a few 
	months back.  I have an additional comment on this topic and a
	change to 
	my previous suggestion:
	
	A search on <idref> within the 1.1 WD gives only two defined
uses
	of the 
	ID property:
	
	1.) as the internal_destination for an fo:basic-link.
	2.) as the ref_id for an fo:page-number-citation.
	
	If this is correct, then given that both defined uses for the
property 
	ID require the user to have advance knowledge of what the ID
value is, 
	the value of subsequently autogenerated ID's for FO's remains
	unclear.  
	OTOH, if autogeneration is primarily for internal processing of
the 
	document, that would appear to be an implementation detail that
doesn't 
	need to be placed in the recommendation.  (FOP, for example,
	currently 
	does not need such an autogenerated ID, and shouldn't be
required to 
	generate one.) ...

Glenn you are correct that the autogenerated ID's are not reachable by
any links (either inside or outside the document). In fact, an
implementation can ignore these IDs. They were put into the
specification to satisfy the following requirements:

1) every property must have an initial value.

2) the "ID" property takes an identifier (string of characters) as its
value and that string must be unique in the result tree.

Requirement 1) means that some value must be specified and requirement
2) means that the value cannot be a keyword, such as "none" because that
would be a legal ID identifier. Furthermore, the value must be unique in
the result tree which eliminates any single initial value.

That lead us to the creation of a random unique identifier that is
unique in the result tree. This is effectively equivalent to your
"implementation dependent value", but more clearly satisfies the above
formal requirements of the specification. Therefore, the Working Group
sees no need to make a change in the specification nor is there any need
for an implementation to realize these not reachable IDs.


        Steve
=====================================
Steve Zilles 
115 Lansberry Court, 
Los Gatos, CA 95032-4710
steve@zilles.org 
Received on Wednesday, 29 September 2004 13:40:23 UTC

This archive was generated by hypermail 2.3.1 : Wednesday, 5 February 2014 07:14:29 UTC