RE: CURIEs, xmlns and bandwidth

Previous Topic Next Topic
 
classic Classic list List threaded Threaded
12 messages Options
Reply | Threaded
Open this post in threaded view
|

RE: CURIEs, xmlns and bandwidth

Misha Wolf

Hi all,
 
We've had confirmation that xmlns declarations can't be XIncluded.
As a consequence, the IPTC does not plan to go the xmlns route for
declaring CURIEs.  If XHTML does decide to use xmlns for CURIE
declarations, that leaves us with two options:
 
1.  Different standards use different declaration mechanisms.
 
2.  The IPTC calls it's CURIEs by some other name.
 
Any comments?

Thanks,
 
Misha Wolf
News Standards Manager, Reuters, www.reuters.com
Vice-Chair, News Architecture Working Party, IPTC, www.iptc.org/dev


-----Original Message-----
From: [hidden email]
[mailto:[hidden email]] On Behalf Of Misha Wolf
Sent: 26 October 2005 06:36
To: [hidden email]
Subject: CURIEs, xmlns and bandwidth


Hi Mark and others,
 
We've been spending a lot of time discussing CURIEs at the IPTC
meeting taking place this week in Milan and were very glad to see
the CURIE Note and the updated RDF/A document.  We have one big
problem, though, and that is this ...
 
We plan to use CURIEs for everything, including the principal
taxonomies themselves, as well as all the related info, like @role,
@creator, @type, etc.
 
If you go to:
  http://www.iptc.org/NewsCodes/
you will see that we currently have nearly 30 taxonomies for things
like Colorspace, Genre, Of Interest To (ie Audience), Provider,
Scene, Status, Subject, Topic Type, Videocoder, ...
 
So a news story with even a moderate number of codes will easily
have 20 or more prefix declarations.  If we assume an average size
of 50 bytes, that makes 1,000 bytes of prefix declarations.  And
this is probably quite a conservative estimate.
 
As we are very concerned about bandwidth, we have been thinking of
allowing the use of XInclude to place the prefix declarations
outside of the news story.  Now we think (possibly wrongly) that
XInclude may not permit the inclusion of arbitrary snippets of
text, at any point in the XML document, eg the inclusion of the
string:
 
  xmlns:nc="http://www.iptc.org/NewsCodes#"
 
at the point shown with "^" below:
 
  <news:item  >
             ^
Having skimmed the (quite dense) XInclude spec, we're still not
quite sure.
 
Prior to seeing the CURIE Note, we had decided to borrow a prefix
declaration syntax from Schematron, eg (these are made-up examples):
 
  <ns prefix="nc" uri="http://www.iptc.org/NewsCodes#"/>
  <ns prefix="lang" uri="http://www.isi.edu/in-notes/bcp/bcp47.txt#"/>
  <ns prefix="curr" uri="http://en.wikipedia.org/wiki/ISO_4217#"/>
 
We believe that XInclude would not get indigestion when presented
with the above.
 
Any comments on the above would be very much appreciated.
 
Thanks,
Misha


-----------------------------------------------------------------
        Visit our Internet site at http://www.reuters.com

To find out more about Reuters Products and Services visit http://www.reuters.com/productinfo 

Any views expressed in this message are those of  the  individual
sender,  except  where  the sender specifically states them to be
the views of Reuters Ltd.


Reply | Threaded
Open this post in threaded view
|

RE: CURIEs, xmlns and bandwidth

Mark Birbeck

Hi Misha,

I haven't thought about how it might work, but there is also another avenue
worth exploring:

3. Devise some other inclusion mechanism, which can include attributes.

Regards,

Mark

 
Mark Birbeck
CEO
x-port.net Ltd.

e: [hidden email]
t: +44 (0) 20 7689 9232
w: http://www.formsPlayer.com/

Download our XForms processor from
http://www.formsPlayer.com/ 

-----Original Message-----
From: [hidden email]
[mailto:[hidden email]] On Behalf Of Misha Wolf
Sent: 28 October 2005 14:51
To: [hidden email]
Cc: [hidden email]
Subject: RE: CURIEs, xmlns and bandwidth


Hi all,
 
We've had confirmation that xmlns declarations can't be XIncluded.
As a consequence, the IPTC does not plan to go the xmlns route for
declaring CURIEs.  If XHTML does decide to use xmlns for CURIE
declarations, that leaves us with two options:
 
1.  Different standards use different declaration mechanisms.
 
2.  The IPTC calls it's CURIEs by some other name.
 
Any comments?

Thanks,
 
Misha Wolf
News Standards Manager, Reuters, www.reuters.com
Vice-Chair, News Architecture Working Party, IPTC, www.iptc.org/dev


-----Original Message-----
From: [hidden email]
[mailto:[hidden email]] On Behalf Of Misha Wolf
Sent: 26 October 2005 06:36
To: [hidden email]
Subject: CURIEs, xmlns and bandwidth


Hi Mark and others,
 
We've been spending a lot of time discussing CURIEs at the IPTC
meeting taking place this week in Milan and were very glad to see
the CURIE Note and the updated RDF/A document.  We have one big
problem, though, and that is this ...
 
We plan to use CURIEs for everything, including the principal
taxonomies themselves, as well as all the related info, like @role,
@creator, @type, etc.
 
If you go to:
  http://www.iptc.org/NewsCodes/
you will see that we currently have nearly 30 taxonomies for things
like Colorspace, Genre, Of Interest To (ie Audience), Provider,
Scene, Status, Subject, Topic Type, Videocoder, ...
 
So a news story with even a moderate number of codes will easily
have 20 or more prefix declarations.  If we assume an average size
of 50 bytes, that makes 1,000 bytes of prefix declarations.  And
this is probably quite a conservative estimate.
 
As we are very concerned about bandwidth, we have been thinking of
allowing the use of XInclude to place the prefix declarations
outside of the news story.  Now we think (possibly wrongly) that
XInclude may not permit the inclusion of arbitrary snippets of
text, at any point in the XML document, eg the inclusion of the
string:
 
  xmlns:nc="http://www.iptc.org/NewsCodes#"
 
at the point shown with "^" below:
 
  <news:item  >
             ^
Having skimmed the (quite dense) XInclude spec, we're still not
quite sure.
 
Prior to seeing the CURIE Note, we had decided to borrow a prefix
declaration syntax from Schematron, eg (these are made-up examples):
 
  <ns prefix="nc" uri="http://www.iptc.org/NewsCodes#"/>
  <ns prefix="lang" uri="http://www.isi.edu/in-notes/bcp/bcp47.txt#"/>
  <ns prefix="curr" uri="http://en.wikipedia.org/wiki/ISO_4217#"/>
 
We believe that XInclude would not get indigestion when presented
with the above.
 
Any comments on the above would be very much appreciated.
 
Thanks,
Misha


-----------------------------------------------------------------
        Visit our Internet site at http://www.reuters.com

To find out more about Reuters Products and Services visit
http://www.reuters.com/productinfo 

Any views expressed in this message are those of  the  individual
sender,  except  where  the sender specifically states them to be
the views of Reuters Ltd.





Reply | Threaded
Open this post in threaded view
|

RE: CURIEs, xmlns and bandwidth

Misha Wolf
In reply to this post by Misha Wolf

Hi Mark,

What are the advantages of using xmlns declarations for CURIEs?  
It is difficut to argue that this approach results in support by
existing tools, as one then has to face the possibility that some
of these tools will get indigestion when faced with CURIEs which
are not legal QNames.

As xmlns and QNames are defined by the Namespaces in XML spec, and
as we're not adopting QNames, why should we use xmlns to declare
constructs which are not QNames?

Regards,

Misha Wolf
News Standards Manager, Reuters, www.reuters.com
Vice-Chair, News Architecture Working Party, IPTC, www.iptc.org/dev


-----Original Message-----
From: Mark Birbeck [mailto:[hidden email]]
Sent: 28 October 2005 15:58
To: Misha Wolf
Cc: [hidden email]; [hidden email]
Subject: RE: CURIEs, xmlns and bandwidth

Hi Misha,

I haven't thought about how it might work, but there is also another
avenue
worth exploring:

3. Devise some other inclusion mechanism, which can include attributes.

Regards,

Mark

 
Mark Birbeck
CEO
x-port.net Ltd.

e: [hidden email]
t: +44 (0) 20 7689 9232
w: http://www.formsPlayer.com/

Download our XForms processor from
http://www.formsPlayer.com/ 

-----Original Message-----
From: [hidden email]
[mailto:[hidden email]] On Behalf Of Misha Wolf
Sent: 28 October 2005 14:51
To: [hidden email]
Cc: [hidden email]
Subject: RE: CURIEs, xmlns and bandwidth


Hi all,
 
We've had confirmation that xmlns declarations can't be XIncluded.
As a consequence, the IPTC does not plan to go the xmlns route for
declaring CURIEs.  If XHTML does decide to use xmlns for CURIE
declarations, that leaves us with two options:
 
1.  Different standards use different declaration mechanisms.
 
2.  The IPTC calls it's CURIEs by some other name.
 
Any comments?

Thanks,
 
Misha Wolf
News Standards Manager, Reuters, www.reuters.com
Vice-Chair, News Architecture Working Party, IPTC, www.iptc.org/dev


-----Original Message-----
From: [hidden email]
[mailto:[hidden email]] On Behalf Of Misha Wolf
Sent: 26 October 2005 06:36
To: [hidden email]
Subject: CURIEs, xmlns and bandwidth


Hi Mark and others,
 
We've been spending a lot of time discussing CURIEs at the IPTC
meeting taking place this week in Milan and were very glad to see
the CURIE Note and the updated RDF/A document.  We have one big
problem, though, and that is this ...
 
We plan to use CURIEs for everything, including the principal
taxonomies themselves, as well as all the related info, like @role,
@creator, @type, etc.
 
If you go to:
  http://www.iptc.org/NewsCodes/
you will see that we currently have nearly 30 taxonomies for things
like Colorspace, Genre, Of Interest To (ie Audience), Provider,
Scene, Status, Subject, Topic Type, Videocoder, ...
 
So a news story with even a moderate number of codes will easily
have 20 or more prefix declarations.  If we assume an average size
of 50 bytes, that makes 1,000 bytes of prefix declarations.  And
this is probably quite a conservative estimate.
 
As we are very concerned about bandwidth, we have been thinking of
allowing the use of XInclude to place the prefix declarations
outside of the news story.  Now we think (possibly wrongly) that
XInclude may not permit the inclusion of arbitrary snippets of
text, at any point in the XML document, eg the inclusion of the
string:
 
  xmlns:nc="http://www.iptc.org/NewsCodes#"
 
at the point shown with "^" below:
 
  <news:item  >
             ^
Having skimmed the (quite dense) XInclude spec, we're still not
quite sure.
 
Prior to seeing the CURIE Note, we had decided to borrow a prefix
declaration syntax from Schematron, eg (these are made-up examples):
 
  <ns prefix="nc" uri="http://www.iptc.org/NewsCodes#"/>
  <ns prefix="lang" uri="http://www.isi.edu/in-notes/bcp/bcp47.txt#"/>
  <ns prefix="curr" uri="http://en.wikipedia.org/wiki/ISO_4217#"/>
 
We believe that XInclude would not get indigestion when presented
with the above.
 
Any comments on the above would be very much appreciated.
 
Thanks,
Misha


-----------------------------------------------------------------
        Visit our Internet site at http://www.reuters.com

To find out more about Reuters Products and Services visit http://www.reuters.com/productinfo 

Any views expressed in this message are those of  the  individual
sender,  except  where  the sender specifically states them to be
the views of Reuters Ltd.


Reply | Threaded
Open this post in threaded view
|

RE: CURIEs, xmlns and bandwidth

Dan Connolly
In reply to this post by Misha Wolf

On Fri, 2005-10-28 at 14:51 +0100, Misha Wolf wrote:

> Hi all,
>  
> We've had confirmation that xmlns declarations can't be XIncluded.
> As a consequence, the IPTC does not plan to go the xmlns route for
> declaring CURIEs.  If XHTML does decide to use xmlns for CURIE
> declarations, that leaves us with two options:
>  
> 1.  Different standards use different declaration mechanisms.
>  
> 2.  The IPTC calls it's CURIEs by some other name.
>  
> Any comments?

I have lost track of many of the relevant constraints and requirements,
but have you considered a sort of microformat approach, using GRDDL?
i.e.

 start with XHTML 1.x
 choose a profile URI; put it in the head of each of your documents
 at that profile URI, add a profileTransformation link to
  a bit of XSLT that expands your compact syntax to full RDF/XML

By way of precedent, let's look at "Embedded RDF"...

Here's an example document
  http://research.talis.com/2005/erdf/foaf-in-html.html

it uses an XHTML dialect for data that Ian Davis designed.

At the top, it declares that dialect/profile:
  <head profile="http://purl.org/NET/erdf/profile">

If we go to that profile document, we see it has
RDF embedded in such a way as to declare a transformation
of all documents that use this profile...
  <head profile="http://www.w3.org/2003/g/data-view">
...
     <link rel="transformation"
       href="http://www.w3.org/2003/g/glean-profile" />
...
  <a
        rel="profileTransformation"
        href="http://purl.org/NET/erdf/extract-rdf.xsl">extract-rdf.xsl</a>

The net result is that we can take foaf-in-html.html and
put it into a GRDDL client (e.g. the online GRDDL service
http://www.w3.org/2003/11/rdf-in-xhtml-demo )
and out comes the data from the page in RDF/XML...

...
   <rdf:Description
rdf:about="http://research.talis.com/2005/erdf/foaf-in-html.html#ian">
       <name xmlns="http://xmlns.com/foaf/0.1/">Ian Davis</name>
       <firstName xmlns="http://xmlns.com/foaf/0.1/">Ian</firstName>
       <surname xmlns="http://xmlns.com/foaf/0.1/">Davis</surname>
...


--
Dan Connolly, W3C http://www.w3.org/People/Connolly/
D3C2 887B 0F92 6005 C541  0875 0F91 96DE 6E52 C29E


Reply | Threaded
Open this post in threaded view
|

RE: CURIEs, xmlns and bandwidth

Misha Wolf
In reply to this post by Misha Wolf

Hi Dan,

We fully intend to use GRDDL to convert the metadata in News Items
to triples.  We've decided to use URIs (expressed as CURIEs) for
*every* term drawn from a vocabulary.  As any individual News Item
will employ many vocabularies, this will require many prefix->URI
declarations.  And we can't afford the impact of using xmlns for
this purpose.  Consider a broadcast stream of real-time headlines.
Let's say that the text of each headline requires 50 bytes.  Let's
also say that the story metadata (which needs to be carried with
the headline to allow filtering by recipients) requires 20
vocabularies and that each prefix declaration takes 50 bytes.  So
having started with 50 bytes of text, we now end up broadcasting
21 * 50 bytes.  This is why we want to use XInclude to allow the
prefix->URI declarations to be outside the headline object.  And
XInclude can't be used for xmlns declarations.

Misha Wolf
News Standards Manager, Reuters, www.reuters.com
Vice-Chair, News Architecture Working Party, IPTC, www.iptc.org/dev


-----Original Message-----
From: Dan Connolly [mailto:[hidden email]]
Sent: 28 October 2005 16:48
To: Misha Wolf
Cc: [hidden email]; [hidden email]
Subject: RE: CURIEs, xmlns and bandwidth

On Fri, 2005-10-28 at 14:51 +0100, Misha Wolf wrote:

> Hi all,
>  
> We've had confirmation that xmlns declarations can't be XIncluded.
> As a consequence, the IPTC does not plan to go the xmlns route for
> declaring CURIEs.  If XHTML does decide to use xmlns for CURIE
> declarations, that leaves us with two options:
>  
> 1.  Different standards use different declaration mechanisms.
>  
> 2.  The IPTC calls it's CURIEs by some other name.
>  
> Any comments?

I have lost track of many of the relevant constraints and requirements,
but have you considered a sort of microformat approach, using GRDDL?
i.e.

 start with XHTML 1.x
 choose a profile URI; put it in the head of each of your documents
 at that profile URI, add a profileTransformation link to
  a bit of XSLT that expands your compact syntax to full RDF/XML

By way of precedent, let's look at "Embedded RDF"...

Here's an example document
  http://research.talis.com/2005/erdf/foaf-in-html.html

it uses an XHTML dialect for data that Ian Davis designed.

At the top, it declares that dialect/profile:
  <head profile="http://purl.org/NET/erdf/profile">

If we go to that profile document, we see it has
RDF embedded in such a way as to declare a transformation
of all documents that use this profile...
  <head profile="http://www.w3.org/2003/g/data-view">
...
     <link rel="transformation"
       href="http://www.w3.org/2003/g/glean-profile" />
...
  <a
        rel="profileTransformation"
 
href="http://purl.org/NET/erdf/extract-rdf.xsl">extract-rdf.xsl</a>

The net result is that we can take foaf-in-html.html and
put it into a GRDDL client (e.g. the online GRDDL service
http://www.w3.org/2003/11/rdf-in-xhtml-demo )
and out comes the data from the page in RDF/XML...

...
   <rdf:Description
rdf:about="http://research.talis.com/2005/erdf/foaf-in-html.html#ian">
       <name xmlns="http://xmlns.com/foaf/0.1/">Ian Davis</name>
       <firstName xmlns="http://xmlns.com/foaf/0.1/">Ian</firstName>
       <surname xmlns="http://xmlns.com/foaf/0.1/">Davis</surname>
...


--
Dan Connolly, W3C http://www.w3.org/People/Connolly/
D3C2 887B 0F92 6005 C541  0875 0F91 96DE 6E52 C29E



-----------------------------------------------------------------
        Visit our Internet site at http://www.reuters.com

To find out more about Reuters Products and Services visit http://www.reuters.com/productinfo 

Any views expressed in this message are those of  the  individual
sender,  except  where  the sender specifically states them to be
the views of Reuters Ltd.


Reply | Threaded
Open this post in threaded view
|

RE: CURIEs, xmlns and bandwidth

Dan Connolly

On Fri, 2005-10-28 at 17:08 +0100, Misha Wolf wrote:
> Hi Dan,
>
> We fully intend to use GRDDL to convert the metadata in News Items
> to triples.  We've decided to use URIs (expressed as CURIEs) for
> *every* term drawn from a vocabulary.  As any individual News Item
> will employ many vocabularies, this will require many prefix->URI
> declarations.  And we can't afford the impact of using xmlns for
> this purpose.

You can just choose "well-known" prefixes and hard-code them
in your XSLT transformation. i.e. well-known to everybody
that uses your profile. Then they don't have to be declared
in each document.

>   Consider a broadcast stream of real-time headlines.
> Let's say that the text of each headline requires 50 bytes.  Let's
> also say that the story metadata (which needs to be carried with
> the headline to allow filtering by recipients) requires 20
> vocabularies and that each prefix declaration takes 50 bytes.  So
> having started with 50 bytes of text, we now end up broadcasting
> 21 * 50 bytes.  This is why we want to use XInclude to allow the
> prefix->URI declarations to be outside the headline object.  And
> XInclude can't be used for xmlns declarations.

Are you broadcasting the headlines in little XHTML documents?
Or in a custom XML vocabulary? Do you have some examples that
you're kicking around? Sorry if I'm asking you to
repeat yourself.

If you help me understand your target, I might be able to
flesh out what I'm suggesting.

--
Dan Connolly, W3C http://www.w3.org/People/Connolly/
D3C2 887B 0F92 6005 C541  0875 0F91 96DE 6E52 C29E


Reply | Threaded
Open this post in threaded view
|

RE: CURIEs, xmlns and bandwidth

Jeremy Carroll


Misha

might it be possible to use a custom DTD to add a fixed set of xmlns
declarations as default attributes.

This could probably be done in a way that would, with the new XHTML2
spec, be standards compliant.

Jeremy





Reply | Threaded
Open this post in threaded view
|

RE: CURIEs, xmlns and bandwidth

Mark Birbeck
In reply to this post by Misha Wolf

Hi Misha,

> What are the advantages of using xmlns declarations for CURIEs?

None, really. It was simply that we began with QNames and then evolved from
that.


> As xmlns and QNames are defined by the Namespaces in XML spec, and
> as we're not adopting QNames, why should we use xmlns to declare
> constructs which are not QNames?

Sure...and going that route would allow you to add some additional cunning
features. Also, in many of the contexts that I have referred to as 'existing
practice' there are no namespaces anyway (software configuration files, Wiki
shortcuts, and so on).

Following on from your comments then, perhaps we should just say that CURIEs
are a datatype, but not say anything about where the actual substitutions
come from, and let the host language or software context decide that. (So
XHTML 2 might choose namespace prefixes, but NewsML might choose some other
way.)

Any thoughts on that?

Regards,

Mark


Mark Birbeck
CEO
x-port.net Ltd.

e: [hidden email]
t: +44 (0) 20 7689 9232
w: http://www.formsPlayer.com/

Download our XForms processor from
http://www.formsPlayer.com/



Reply | Threaded
Open this post in threaded view
|

RE: CURIEs, xmlns and bandwidth

Misha Wolf
In reply to this post by Misha Wolf

Hi Dan,

> > We fully intend to use GRDDL to convert the metadata in News
> > Items to triples.  We've decided to use URIs (expressed as
> > CURIEs) for *every* term drawn from a vocabulary.  As any
> > individual News Item will employ many vocabularies, this will
> > require many prefix->URI declarations.  And we can't afford
> > the impact of using xmlns for this purpose.

> You can just choose "well-known" prefixes and hard-code them
> in your XSLT transformation. i.e. well-known to everybody
> that uses your profile. Then they don't have to be declared
> in each document.

Within our standards, each news provider is free to use their own
taxonomies, eg of subjects, entities, genres, etc.  So there are
no "well-known" prefixes to hardwire.

> > Consider a broadcast stream of real-time headlines.
> > Let's say that the text of each headline requires 50 bytes.  
> > Let's also say that the story metadata (which needs to be
> > carried with the headline to allow filtering by recipients)
> > requires 20 vocabularies and that each prefix declaration
> > takes 50 bytes.  So having started with 50 bytes of text, we
> > now end up broadcasting 21 * 50 bytes.  This is why we want to
> > use XInclude to allow the prefix->URI declarations to be
> > outside the headline object.  And XInclude can't be used for
> > xmlns declarations.

> Are you broadcasting the headlines in little XHTML documents?
> Or in a custom XML vocabulary?

NewsML 2 (under development) uses our own XML Schema for the
metadata (NewsML 1 is DTD-based).  NewsML is content-agnostic, and
may carry any payload.  Where the payload is text, some IPTC
members (including Reuters) use XHTML, while others use a markup
language called NITF.

> Do you have some examples that you're kicking around? Sorry if
> I'm asking you to repeat yourself.

Our current draft looks like this.  The URI examples are just that
(ie examples).

<news:item>
  <dsig:Signature/>
  <catalog>
    <ns prefix="nc" uri="http://www.iptc.org/NewsCodes#"/>
    <ns prefix="rtr" uri="http://www.reuters.com/NewsCodes#"/>
    <ns prefix="lang" uri="http://www.isi.edu/in-notes/bcp/bcp47.txt#"/>
    <ns prefix="curr" uri="http://en.wikipedia.org/wiki/ISO_4217#"/>
    ...
  </catalog>
  <itemMeta>
    ...
  </itemMeta>
  <contentMeta>
    <created>2005-10-23T12:34:56Z</created>
    <creator code="afp:llm"/>
    <contributor code="greekMythology:muse"/>
    <source code="org:iptc"/>
    <significance>100</significance>
    <audience code="aud:implementors"/>
    <service code="service:tech"/>
    <edNote>Eat afer reading</edNote>
    <title>Hello World</title>
    <description>Something or other ...</description>
    <subject code="nc:04008018"/>
    <subject code="rtr:123"/>
    <subject code="rtr:456"/>
    <subject code="rtr:789"/>
    <subject code="curr:JPY"/>
    <genre code="spec:tech"/>
    <language code="lang:zh-Hant"/>
    ...
  </contentMeta>
  <news:content>
    ...
  </news:content>
</news:item>

> If you help me understand your target, I might be able to
> flesh out what I'm suggesting.

Thanks,

Misha Wolf
News Standards Manager, Reuters, www.reuters.com
Vice-Chair, News Architecture Working Party, IPTC, www.iptc.org/dev


-----------------------------------------------------------------
        Visit our Internet site at http://www.reuters.com

To find out more about Reuters Products and Services visit http://www.reuters.com/productinfo 

Any views expressed in this message are those of  the  individual
sender,  except  where  the sender specifically states them to be
the views of Reuters Ltd.


Reply | Threaded
Open this post in threaded view
|

RE: CURIEs, xmlns and bandwidth

Misha Wolf
In reply to this post by Misha Wolf

Hi Mark,

> > What are the advantages of using xmlns declarations for CURIEs?

> None, really. It was simply that we began with QNames and then
> evolved from that.

> > As xmlns and QNames are defined by the Namespaces in XML spec,
> > and as we're not adopting QNames, why should we use xmlns to
> > declare constructs which are not QNames?

> Sure...and going that route would allow you to add some
> additional cunning features.

Indeed.  The advantages of using the Schematron syntax rather than
the xmlns syntax for prefix declarations include:

-  The ability to XInclude the prefix declarations.

-  The ability to group the prefix declarations by taxonomy
   provider, and XInclude such groups of declarations separately.

-  No problem of tools telling us that these (non-QNames) aren't
   legal QNames.

-  The ability to add additional attributes to the declarations,
   should this become useful.  It may be that RDDL will suffice for
   adding information about schemes (our term for taxonomies), but
   having the ability to add attributes should these become
   necessary is good insurance.

> Also, in many of the contexts that I have referred to as
> 'existing practice' there are no namespaces anyway (software
> configuration files, Wiki shortcuts, and so on).
>
> Following on from your comments then, perhaps we should just say
> that CURIEs are a datatype, but not say anything about where the
> actual substitutions come from, and let the host language or
> software context decide that. (So XHTML 2 might choose namespace
> prefixes, but NewsML might choose some other way.)
>
> Any thoughts on that?

As we see it, there is no deployed technology that satisfies the
requirements.  So the IPTC will use CURIEs regradless of their
adoption by the W3C.

The ideal outcome would be for CURIEs to be used in a uniform way
across the Web, with a single declaration syntax.

A less-good outcome would be for CURIEs to be used in a uniform way
across the Web, with more than one declaration syntax.

The worst outcome would be for there to be multiple solutions to
the overall problem.

Regards,

Misha Wolf
News Standards Manager, Reuters, www.reuters.com
Vice-Chair, News Architecture Working Party, IPTC, www.iptc.org/dev


-----------------------------------------------------------------
        Visit our Internet site at http://www.reuters.com

To find out more about Reuters Products and Services visit http://www.reuters.com/productinfo 

Any views expressed in this message are those of  the  individual
sender,  except  where  the sender specifically states them to be
the views of Reuters Ltd.


Reply | Threaded
Open this post in threaded view
|

RE: CURIEs, xmlns and bandwidth

Dan Connolly
In reply to this post by Misha Wolf

On Mon, 2005-10-31 at 13:43 +0000, Misha Wolf wrote:
[...]
> Within our standards, each news provider is free to use their own
> taxonomies, eg of subjects, entities, genres, etc.  So there are
> no "well-known" prefixes to hardwire.

Oh. Well, then you'll have to transmit the full URIs somehow.
I got the impression that the xmlns declarations introduced
a lot of redundancy. But if the taxonomies are completely
unconstrained, then I don't see how you'll do substantially
better than xmlns declarations.


--
Dan Connolly, W3C http://www.w3.org/People/Connolly/
D3C2 887B 0F92 6005 C541  0875 0F91 96DE 6E52 C29E


Reply | Threaded
Open this post in threaded view
|

RE: CURIEs, xmlns and bandwidth

Misha Wolf
In reply to this post by Misha Wolf

By XIncluding the declarations :-)

Let's say I send 10,000 headlines, all XIncluding the
same set of declarations.  That allows me to transport
the declarations just once, rather than 10,000 times.

Misha


-----Original Message-----
From: Dan Connolly [mailto:[hidden email]]
Sent: 31 October 2005 14:26
To: Misha Wolf
Cc: [hidden email]; [hidden email];
[hidden email]
Subject: RE: CURIEs, xmlns and bandwidth

On Mon, 2005-10-31 at 13:43 +0000, Misha Wolf wrote:
[...]
> Within our standards, each news provider is free to use their own
> taxonomies, eg of subjects, entities, genres, etc.  So there are
> no "well-known" prefixes to hardwire.

Oh. Well, then you'll have to transmit the full URIs somehow.
I got the impression that the xmlns declarations introduced
a lot of redundancy. But if the taxonomies are completely
unconstrained, then I don't see how you'll do substantially
better than xmlns declarations.


--
Dan Connolly, W3C http://www.w3.org/People/Connolly/
D3C2 887B 0F92 6005 C541  0875 0F91 96DE 6E52 C29E



-----------------------------------------------------------------
        Visit our Internet site at http://www.reuters.com

To find out more about Reuters Products and Services visit http://www.reuters.com/productinfo 

Any views expressed in this message are those of  the  individual
sender,  except  where  the sender specifically states them to be
the views of Reuters Ltd.