Re: proposals for Lists and Seq (ISSUE-77)

On 20 Oct 2011, at 14:39, Sandro Hawke wrote:

> On Thu, 2011-10-20 at 14:58 +0200, Dan Brickley wrote:
>> On 20 October 2011 14:47, Sandro Hawke <sandro@w3.org> wrote:
>>> On Thu, 2011-10-20 at 13:21 +0200, Dan Brickley wrote:
>>>> On 20 October 2011 13:13, Steve Harris <steve.harris@garlik.com> wrote:
>>>>> I wouldn't be comfortable with marking Seq as "archaic" or similar unless there's a viable alternative, and I don't think List counts.
>>>> 
>>>> Me neither.  Nor "quaint", "twee", "retro" or "regrettable". It's just
>>>> what it is, with no great mystery or confusion.
>>> 
>>> Actually, there's a great deal of confusion.  Please do explain -- in
>>> one sentence for newbies -- why we have both Seq and List, and with Seq
>>> better supported in RDF/XML and List better supported in Turtle, and how
>>> someone should decide which to use.
>> 
>> Sure - good idea.
>> 
>> "The first RDF/XML specifications used a class rdf:Seq with numbered
>> relationships to describe ordered lists; however when the later OWL WG
>> were arm-twisted by W3C staff into using RDF as the syntax to define
>> their language, they persuaded the RDF Core group (and others, e.g.
>> N3/Turtle) to adopt a new list mechanism that used a linked list style
>> that made it easier to tell when a list description was incomplete, at
>> the cost of extra triples."
> 
> Nice :-) but you didn't answer the second-half of the question.
> 
> Also, why didn't you (the 2001-2004 RDF Core WG) just add an end-marker,
> if that was the objective?     I missed that particular debate.
> 
>>> (To put it differently, I think it's quite harmful to the RDF to not deprecate Seq.)
>> 
>> I guess you're not paying by the triple for storage?
> 
> I should hope that no one stores well-formed lists as triples.  

I've never worked with a system that doesn't store them as triples, though there might be some. Lists are used so little that the amount of engineering effort required to support them efficiently is not justified. It's a really significant problem to synthesise the triples from a more efficient internal array storage implementation.

> Can you think of any way we could merge Seq and List?   Have well-formed
> sequences and well-formed lists both mappable to native lists?   That
> would work fine, except for round-tripping….

I gave this some thought when you brought it up before (well, mapping Lists to native data structures) and it seems difficult.

- Steve

Received on Thursday, 20 October 2011 14:26:13 UTC