RE: imports and absolute URIs

Hello Jeremy,

I like your description of the problem and agree with most of your conclusions apart from the last one: I believe that this is quite
a common task that people are trying to accomplish, it is related to the language, and therefore should be solved by in this WG.
I've experienced first hand what happens when users don't understand how these issues are to be handled: they usually get confused,
frustrated, and create ontologies that contain errors, which in the end hampers interoperability.

I believe that this example you've given verily demonstrates that the problem of imports is related to the Web and therefore is in
the scope of the WG. As you've pointed out, the makers of HTML have correctly identified the need to have some sort of relative
addressing scheme. Moreover, what makes the Web different from my file folder structure? Both have a URI scheme for identifying
resources and have some protocol for accessing the files. The problems that you'd have with moving OWL files on your local folders
are the same as the problems that you'd get if you wanted to reorganize them on your Web server. Having some mechanism for locating
relevant files independent of their location, either on the servers or locally, would solve this problem. Furthermore, it exists in
HTML and was in the scope of W3C; hence, I can't see why the same would not hold for OWL.

Regards,

	Boris

> -----Original Message-----
> From: public-owl-wg-request@w3.org [mailto:public-owl-wg-request@w3.org] On Behalf Of Jeremy Carroll
> Sent: 21 December 2007 17:08
> To: Web Ontology Language ((OWL)) Working Group WG
> Subject: imports and absolute URIs
> 
> 
> 
> I keep running the following analogy in my head, and it doesn't quite
> work, which does show that the imports issue has a little more content
> than I have been making out ...
> 
> Analogy
> =======
> 
> Xiao-Wei wants to publish a Web page in HTML, with a picture of his
> family, as a PNG image.
> 
> He gets out his camera and takes the photo, transfers the image to his
> PC, converts the format.
> 
> He then starts up his text editor, writes his html page, and adds
>    <img src="family.png">
> and looks at the page in his Web browser (looking at #
> file://localhost/tmp/mypage.html
> )
> when he is satisfied he uploads both files (the HTML file, and the PNG
> file) to a Web server,at example.org, and everything works (looking at
> http://www.example.org/xiao-wei/mypage.html
> )
> 
> He does the same with his ontology files, and it doesn't work.
> 
> Returning to the HTML, PNG files,
> example.org start to recommend that he uses the html base element to
> declare a base URI for his HTML files. He does this, and he finds that
> his previous edit/preview cycle stops working, because the preview of
> the local HTML file uses the image file from the Web, rather than from
> his local machine.
> 
> -----
> 
> Carefully thinking through when the ontology edit/upload stops working,
> it is with the addition of an xml:base, or when we refer to an entity
> defined in one of the ontologies using an absolute URI (typical through
> a namespace).
> 
> This is required (rather than using local URIs) because of the decision
> that XML Namespaces must be absolute URIs and not local ones.
> 
> Which is the topic of another current thread
> see e.g.
> 
> http://lists.w3.org/Archives/Public/semantic-web/2007Dec/0117
> 
> ====
> 
> There is a mess here, in how to move a group of files that reference
> each other using absolute references from location A to location B on
> the web.
> 
> OWL imports is an instance of this problem, because while the imports
> can and should be done using relative refs, at various points absolute
> references are needed.
> 
> I don't believe we should be solving that problem in this group.
> 
> I think a good practice, for low-tech tools, would be to have near the
> top of each of the files a single XML entity definition that gives the
> 'top' of the logical hierarchy. As you move the docs around, you would
> need to change that.
> 
> I would expect many tools would want to be more sophisticated.
> 
> 
> Jeremy
> 
>

Received on Friday, 21 December 2007 21:58:52 UTC