RE: [xml-dev] Re: determining ID-ness in XML

["Bullard, Claude L (Len)" <clbullar@ingr.com>]

PIs or xml vocabulary.  One can be ignored now, the 
other has to be acknowledged by fiat.   It depends 
I guess on how reliable we want this to be.  

ID resolution in this case should be reliable.  That 
argues for extending the system vocabulary instead 
of PIs or we would have to elevate a class of PIs 
for special treatment.  Seems to me, the attribute 
solution would be best.

Again:  requirements.  I'm not arguing against IDs: 
I want requirements up front.  Otherwise, XML is a 
bad hack.

If it takes 20 years, we are idiots.

len

-----Original Message-----
From: Christopher R. Maden [mailto:crism@maden.org]

At 06:18 2-11-2001, Bullard, Claude L (Len) wrote:
>The PI solution is the least impact solution so far.  Where
>one comes down on this seems to depend on how one thinks
>the system or processor vocabulary should be extended and
>what an XML processor must enforce.  Again, means have always
>been there and by the circular logic that seems to pervade
>the W3C thinking, people are free to ignore those means.
>So we add PIs.  They can ignore those too.  So we add
>more system vocabulary attributes; they can ignore those
>too.  Will we need Congress to legislate this stuff too?

Systems are free to ignore the PIs.  Things will break - this is TimBL's 
great realization that made the Web possible.  Things break.  Let them.

But there is a very clearly perceived benefit to not parsing external 
resources - whether valid or not, there are significant processing models 
built this way.  Putting the ID PIs into the document instance at least 
makes it possible to detect IDs without external resources.  Without an 
in-instance identifier, any reliability of ID-based addressing goes out the 
window.  No other datatype presents this problem; you can cast, you can 
account for it, you can ignore it, and get reasonable results.  IDs are 
special, and I think they warrant this level of special treatment.

>The bizarreness of this to me Chris, is that these are
>well known issues of hypermedia regardless of the notation.

Well, yeah... we learn a few more lessons each go-around, and in twenty 
years perhaps we'll be having similar debates about stylesheet-based 
high-quality formatting, having solved the markup and linking problems.