Re: [Turtle] Two formats

On 3/5/2011 9:58 AM, Richard Cyganiak wrote:
> Lee,
>
> On 4 Mar 2011, at 20:03, Lee Feigenbaum wrote:
>>> I'd like to hear more about real-world use of TriG.
>>
>> We use TriG for:
>>
>> + small, hand-written configuration files
>> + small, system-generated configuration files
>> + ontology exports
>> + small and large data set exports
>> + bundling applications for deployment between environments
>> + custom RDF generation from 3rd party scripts&  apps for import into Anzo
>> + ...
>>
>> Basically, it's Anzo's format of choice for any serialization of RDF. Given that:
>>
>> a) All RDF within Anzo is within a named graph
>> b) We find TriG to be far more human-friendly than something like N-quads
>
> Right.
>
>> ...we tend to choose it as our default format for just about everything.
>
> Ok, thanks for the details. I see why you definitely want to see a TriG-like format as one of the results of this standardization process.

Yup.

> I'm still concerned that TriG has been around for seven years, and only seen a handful of users, and most if not all of them seem to use it only as an internal format within a single system. It is not used as an exchange format.

FWIW, I do understand this hesitation.

I do believe that there is a bit of a chicken-and-egg situation here. I 
can identify many occasions where people I work with were not aware of 
TriG and never would have chosen to use it until I taught them about it, 
and I can identify some occasions on which I published RDF (ok, I don't 
do this that often) as triples rather than quads because of the lack of 
a standard quads/named graphs serialization.

All of which is to say, while I'd be the first person to agree that most 
of the time standards groups need to codify existing practice, I do 
believe there is a role for standards groups to identify a gap in 
existing practice and prescribe a standard by choosing among various 
existing alternatives. I respect that not seeing a ton of deployed TriG 
in the wild is a reason to pause, but without an obviously superior 
alternative and without substantial technical objections, it wouldn't be 
enough of a reason for me to believe that it should _not_ be a top 
candidate for named graph serialization at this point.

Of course, I'm admittedly biased. :-)

Lee

>
> Best,
> Richard
>
>
>
>
>>
>> Lee
>>
>>>
>>> Here's what I'm aware of:
>>>
>>> 1. As examples for NG4J/WIQA (this is what we created TriG for)
>>> 2. As a language for expressing spreadsheet-to-RDF mappings in XLWrap
>>> 3. As a syntax for configuration files (?) in your system.
>>> 4. As a “graph store persistence” format in the Semantic Web Client Library and derived systems
>>>
>>> None of these require exchange between different systems. They're all about local storage or configuration. If the use cases for human-written TriG boil down to configuration files, then I'm unconvinced that this WG should put a format on REC track for that.
>>>
>>> N-Quads on the other hand is used quite often to exchange dumps between different parties (DBpedia publishes N-Quads; the Billion Triples dataset is available as N-Quads; Sindice can process N-Quad dumps).
>>>
>>> So I see a clear case for standardizing a multi-graph dump format, but not such a clear case for standardizing a multi-graph “small-scale” format a la SuperTurtle or TriG.
>>>
>>> Best,
>>> Richard
>>>
>>>
>>>
>>>>
>>>> Lee
>>>>
>>>>>
>>>>> Best,
>>>>> Richard
>>>>>
>>>>> [1] http://sw.deri.org/2008/07/n-quads/
>>>>>
>>>>>
>>>>>
>>>>>
>>>>>
>>>>>>
>>>>>> [the rest of your email has good stuff, but I don't have time to respond
>>>>>> at the moment.]
>>>>>>
>>>>>>     -- Sandro
>>>>>>
>>>>>>
>>>>>>
>>>>>>
>>>>>>> If we can't change turtle, and can't do super-turtle or qurtle, why and
>>>>>>> how can we even discuss graphs of any form?
>>>>>>>
>>>>>>> Syntax sugar like ^ prefix, in scope?
>>>>>>>
>>>>>>> Quoted Graphs, in scope?
>>>>>>>   - if yes, what to they resolve to in the RDF semantics? how would that
>>>>>>> work?
>>>>>>>
>>>>>>> Graph Literals?
>>>>>>>   - what's the difference between quoted graphs?
>>>>>>>
>>>>>>> variables?
>>>>>>>
>>>>>>> changes to the semantics?
>>>>>>>   - if no, can changes like g-box be introduced without being in the
>>>>>>> semantics?
>>>>>>>
>>>>>>> changes to the concepts?
>>>>>>>   - if yes, what about B.C. with RDF/XML? existing deployed data and
>>>>>>> processors? how can they change but the semantics not?
>>>>>>>
>>>>>>> align turtle with sparql?
>>>>>>>   - if yes, how without variables, subject literals and all the other bits?
>>>>>>>
>>>>>>> Sorry, I feel like we need to know what definitely cannot happen, what
>>>>>>> definitely can and what's a grey area, for this WG.
>>>>>>>
>>>>>>> Cheers,
>>>>>>>
>>>>>>> Nathan
>>>>>>>
>>>>>>>
>>>>>>
>>>>>>
>>>>>>
>>>>>
>>>>>
>>>>>
>>>
>>>
>>
>
>

Received on Saturday, 5 March 2011 18:07:10 UTC