Re: Problem with auto-generated fragment IDs for graph names

On 2013-02-14, at 12:58, Eric Prud'hommeaux <eric@w3.org> wrote:

> * Steve Harris <steve.harris@garlik.com> [2013-02-14 12:12+0000]
>> On 2013-02-14, at 12:05, William Waites <wwaites@tardis.ed.ac.uk> wrote:
>> 
>>> I don't think having unnamed graphs is at all strange. RDF graphs are
>>> anomalous in that they are the only kind of resource that we can talk
>>> about using persistent global names but can't talk about using
>>> temporary local names. I find this anomaly to be strange.
>>> 
>>> But given Andy's remarks the last time around, I agree that this ship
>>> has sailed for RDF 1.1, we'll have to wait for the next iteration to
>>> fix it.
>> 
>> I think that people who've tried to work around the "default" / "unanamed" graph (note, singular) in SPARQL can agree that's it's at best unfortunate, and at worst a giant pain in the arse.
> 
> I understand why you mention the default graph as a cautionary tale;
> it's always been a weird wart on the edge of an otherwise pretty
> consistent system, leaving us with questions about how to define it,
> how to transfer state, etc.
> 
> I don't, however, think that it really informs us about the complexity
> of bnode-labeled graphs (note, plural). From the perspective of
> writing C code, re-using BNodes for graph labels was exactly zero
> work. There was no point where I felt like it was unclear how to
> implement SPARQL or dataset merging or anything else.

I was more thinking from a user p.o.v., I agree that implementing it is easy. 3store had bNode graph labels - for generated graphs with inferred triples in them.

But, it makes referring to those graphs in other data tricky, as you don't know what the lifetime of the identifier/label/whatever will be.

- Steve

>> If you have exactly one graph, then there's no issue (you never need to refer to it), but as soon as you have >1 it starts to bite you.
>> 
>> All this thinking about trying to save bytes in the representation I find very concerning - it seems to be assuming that this data is being typed by someone - if your system has people typing in significant quantities of RDF then something is pretty odd, IMHO. We should be aiming for clarity, and unambiguity IMHO, and I don't see how anonymous graphs help.
>> 
>> - Steve
>> 
>> -- 
>> Steve Harris
>> Experian
>> +44 20 3042 4132
>> Registered in England and Wales 653331 VAT # 887 1335 93
>> 80 Victoria Street, London, SW1E 5JL
>> 
>> 
> 
> -- 
> -ericP
> 

-- 
Steve Harris
Experian
+44 20 3042 4132
Registered in England and Wales 653331 VAT # 887 1335 93
80 Victoria Street, London, SW1E 5JL

Received on Friday, 15 February 2013 11:39:29 UTC