Re: property paths

On Mar 26, 2012, at 5:29 PM, Polleres, Axel wrote:

> Hi Greg, all,
> 
> Indeed, what you write is some of the commenters argument.
> However, last time we agreed on sticking - as the default - 
> with the counting semantics and having a switch to turn on 
> non-counting semantics on full paths (DISTINCT()) rather than 
> the other way around (ALLPATHS()).

Ah. Based on the resolution text, it wasn't clear to me that all of this was discussed.

> Unless you object to last week's decision, I'd prefer not to reopen the 
> issue at this point and move on with the resolution from last week, 
> putting proposals to default path semantics over the whole query to 
> noncounting or having finer grained distinction, such as DISTINCT() 
> at subpath-level to discussion for the future work items list.

I'm not sure about objecting, but I think it's crazy that we'd spec two different path semantics, and then choose to give one an obviously more desirable syntax, and the other a less desirable syntax, when we have (to my knowledge) zero data on how they are going to be used and whether one will turn out to be more widely used than the other.

> We had the issue that common practice required syntax not directly 
> available in the standard already in the previous addition and fixed 
> that now (MINUS). So, I guess it's fair enough to proceed as we have 
> resolved now on propoert paths and see how it's adopted, whereupon a 
> future WG can react, as necessary, as we did with MINUS. 
> Agreed?

No, I don't think it's like the MINUS case at all. In this case, we're providing two alternative semantics for the same feature, and giving one a better syntax. If it turns out that basically everybody wants the DISTINCT semantics by default, and the counting semantics are rarely used, based on our current decision we *can't* come back later in another WG and decide to change the default semantics to align with what people want. We're treating one of the semantics preferentially, and as far as I can tell, it's only because we spec'ed one of them out earlier than the other. Are there other compelling reasons to prefer counting semantics? Things like Jeen's comments suggest to me that for many people it might be the other way around.

I suspect I'm willing to be dragged along on this, but it seems to me very much like we're making some big decisions about the standard without any real-world experience to guide us.

.greg

Received on Monday, 26 March 2012 22:04:31 UTC