Re: Restpark - Minimal RESTful API for querying RDF triples

Hi Kingsley and Luca,

I like Luca's thinking, I don't think it diminishes the role of SPARQL and
Virtuoso. Instead, correct me wrong, this API is Big Data inspired.

The Big Data world is about using MapReduce and its friends to build
parallel indexes that are made available through a key-value store. In this
case we run MapReduce over our harvested Linked Data to build (at least)
three indexes -- for subjects properties and objects. Because the indexes
are build in advance, we get outstanding lookup performance over terabytes
of data.

Google and Amazon tell us we can scale indexes as much as we like [1,2].
However, this is over a high speed network. The question is whether these
techniques adapt to achieve scalability in a higher latency environment, or
whether the performance of the distributed indexes fall flat on their face.
If they do scale maybe Luca is spot on the right track by providing an API
at the index level...

Best regards,

Ross

[1] *Bigtable*: A distributed storage system for structured
data<http://dl.acm.org/citation.cfm?id=1365816>
[2] *Dynamo*: *amazon's *highly available key-value
store<http://www.read.seas.harvard.edu/~kohler/class/cs239-w08/decandia07dynamo.pdf>



On 17 April 2013 01:49, Kingsley Idehen <kidehen@openlinksw.com> wrote:

>  On 4/16/13 3:29 PM, Luca Matteis wrote:
>
> Mark, true this isn't actually REST.
>
>  I just felt that REST is such a buzzword these days for defining simple
> and easily implementable web services. Which is exactly what Restpark aims
> to be.
>
>  Kingsley, it seems like you haven't used databases such as CouchDB that
> *only* expose an HTTP API (this on is actually fully RESTful). CouchDB is
> an entirely scalable and widely used database solution, so I don't fully
> understand your scaling argument.
>
>
> You have basically confirmed my point!
>
> CouchDB has a CouchDB specific collection of interaction patterns. SPARQL
> is one interaction pattern equipped with a powerful declarative query
> language. Remember, this is the Web of Data not the Web of CouchDB
> accessible data etc.. :-)
>
>
> Kingsley
>
>
>  I totally understand that SPARQL is meant to "do it all" and that's
> obviously needed. The purpose of Restpark is not to "do it all" but simply
> to offer the basic features sufficient for many uses cases.
>
>
> On Tue, Apr 16, 2013 at 9:06 PM, Mark Baker <distobj@acm.org> wrote:
>
>> On Tue, Apr 16, 2013 at 1:52 PM, Luca Matteis <lmatteis@gmail.com> wrote:
>> > I have recently created Restpark: http://lmatteis.github.io/restpark/
>> >
>> > It's my way of pushing a standard RESTful interface for accessing RDF
>> data.
>>
>>  Unfortunately, this isn't REST. REST APIs must use hypermedia;
>>
>> http://roy.gbiv.com/untangled/2008/rest-apis-must-be-hypertext-driven
>>
>> A hypermedia solution would declare a form that described how to
>> construct a new URI which could be used to retrieve the data via GET.
>> The form would provide the parameter names (e.g. "subject") along with
>> their types (e.g. rdf:subject).
>>
>> Mark.
>>
>
>
>
> --
>
> Regards,
>
> Kingsley Idehen	
> Founder & CEO
> OpenLink Software
> Company Web: http://www.openlinksw.com
> Personal Weblog: http://www.openlinksw.com/blog/~kidehen
> Twitter/Identi.ca handle: @kidehen
> Google+ Profile: https://plus.google.com/112399767740508618350/about
> LinkedIn Profile: http://www.linkedin.com/in/kidehen
>
>
>
>