Tom Anderson wrote:
> On Fri, 5 Dec 2008, Philippe Poulard wrote:
> 
>> This is a somewhat deprecated usage inherited from the SGML days. I 
>> remember I used such machinery 10 years ago to have documents composed 
>> of text and binary contents. This had some sense since tools were 
>> supported it (I mean SGML/XML editors), [...]
> 
> Ah, i see. Thanks for your answer. I will proceed to completely forget 
> about notations and external entities!

Only if you plan on using XML for data encoding or transmission. If you 
use XML for normal text documents then Notations and Entities (and 
Processing Instructions) are important tools for document management.

 >> but today, people tend to
 >> stick media to their documents à la HTML, with a simple href
 >> attribute.

Only for trivial documents. If you are doing large-scale long-term 
complex document management, simple hrefs just don't cut the mustard.

[tom]
>>> How about notations and external entities? Is the idea that they're a 
>>> mechanism of linkage to external files that's more concrete than just 
>>> using a URL? So, if i was writing a bizarro world HTML, i could 
>>> specify images like this in the DTD:
>>>
>>> <!NOTATION jpeg SYSTEM "http://some-kind-of-URL">
>>> <!NOTATION png SYSTEM "http://some-other-kind-of-URL">
>>> <!ELEMENT img EMPTY>
>>> <!ATTLIST img
>>>     src ENTITY
>>>     alt CDATA #IMPLIED >
>>>
>>> Then in my document i could write:
>>>
>>> <!DOCTYPE img PUBLIC "-//Bizarro HTML" "http://bizarrohtml" [
>>>  <!ENTITY lena SYSTEM "lena.jpg" NDATA jpg>
>>> ]>
>>> <img src="lena" alt="picture of Lena"/>

Yes, exactly, except that the NOTATION declaration can be used (some 
would say abused) for two things:

<!NOTATION jpeg PUBLIC "ISO/IEC 10918-1:1994//NOTATION Digital 
Compression and Coding of Continuous-tone Still Images (JPEG)" SYSTEM 
"/usr/bin/eog">

A formal catalog can be used to detect the FPI and verify that the image 
conforms to the specification; and the SI can be used by a processor to 
run the specified program on the image (eg to embed it in a PDF, or 
convert it to some other format).

>>> Can i also use that entity in regular text, like:
>>>
>>> <p>Here is a picture of Lena &lena;</p>

No, because the processor expects inline entities to resolve to 
processable XML text or markup.

But another reason for using the technique is for repeated images such 
as navigation icons. You really don't want to have to add <icon 
uri="http://some.host.name/dir/dir/dir/someicon.gif"/> every time, 
especially by hand, when <icon imgref="nextchap"/> is simpler and 
easier, and lets you manage the icon file references once at the top of 
the file or in the DTD or (more likely) in a file of entity declarations 
which can be maintained by a non-XML expert or generated from a database.

>>> And in both cases, what does it *mean*? 

The extra level of indirection provides a form of safety-net for your 
documents. If you have a data warehouse with all 35,000 of your books, 
articles, manuals, catalogs, whatever, all referencing obsolete URIs 
that constantly needs updating, you'll find an entity mechanism and a 
single file a much more efficient way to do it.

>>> If i parsed that into DOM and 
>>> called getAttribute("src") on the img element, what would i get back?

I would expect it to return the name of the entity ("lena"). A different 
function should be available (as in XSLT) to resolve the entity 
reference against the declaration and return the name of the physical file.

>>> What's the point of being able to declare an attribute as being of 
>>> type NOTATION?

Explained above.

>>> And does anyone actually use any of this stuff?

Yes, extensively. But typically only in the text document management 
field. Users of rectangular XML (eg spreadsheet data) woud not normally 
have any use for this stuff at all.

///Peter
-- 
XML FAQ: http://xml.silmaril.ie/