Re: ISSUE-94 (n-ary constucts and RDF): Problem with roundtripping when going from functional-style syntax into RDF and back

On 29 Jan 2008, at 14:42, Ian Horrocks wrote:

> On 29 Jan 2008, at 14:28, Bijan Parsia wrote:
>
>>
>> On 29 Jan 2008, at 14:12, Ivan Herman wrote:
>>
>>> Well... You might think that the issue is only for complicated  
>>> URIs like http://a.b.c/?qqq&www (which may indeed be a  
>>> pathological case for a property name) but it is not. We (ie,  
>>> W3C) had long discussions in the past few months with IPTC[1]  
>>> who, for historical reasons, have a bunch of URI-s of the form  
>>> http://a.b.c/123 (ie, with numerals at the end of the URI  
>>> string), but they would like to use RDF & co for, eg, their  
>>> definition of NewsML[2]. Ie, they may have very good reasons to  
>>> use property names of this form in an ontology.
>>>
>>> Ie: I would be cautious in introducing such restriction...
>>
>> I concur,
>
> Still seems bizarre to me that our syntax spec doesn't specify  
> names in a way that ensures that they can be written down using the  
> standard concrete syntax -- or, looked at the other way round, that  
> the standard concrete syntax doesn't allow us to write down all  
> legal names. But what do I know.

I discussed this with the chair of RDF Core after their second last  
call intending to make a LC comment on it, but it seemed that the  
obvious fix wouldn't happen so I dropped it.

Yep, it's a weird thing.

[snip]
>> (And, as I've said long ago, I wouldn't mind tightening some of  
>> the serialization requirements in the RDF/XML (i.e., order of  
>> axioms, etc.) In fact, I think that's highly desirable.)
>
> Doe this impact on round tripping?

I don't *think* so as long as the prettyprinting step doesn't change  
any constructors...which it shouldn't. (I.e., roundtripping is  
Functional syntax to graph; pretty printing is function syntax *or*  
graph to a specific linear form.)

So I think round tripping (RT)  is a component of "nice  
serialization" (NS), so the dependancy runs from RT to NS. Even  
without RT, some specced NS would be a good idea.

Cheers,
Bijan.

Received on Tuesday, 29 January 2008 14:52:03 UTC