As usual you're way ahead of me Reto (not a big compliment, the dogs are too).

Didn't know about the DSL.

But clerezza must tick some boxes for it to be worthwhile: science
project, tick. Blog engine is a tick, but no more a big one, Facebook
won that game. Doing ID well is a good tick, but not if it's just some
random project on github.

I like the raw engine of HTTP, but that can be satisfied by sticking a
4-bit chip up my dog's bum (sorry Sasha).

Reto, I want to find Clerezza so compelling I have no choice. Hoppity.

On 12 May 2011 19:50, Reto Bachmann-Gmuer <[email protected]> wrote:
> On Thu, May 12, 2011 at 7:16 PM, Danny Ayers <[email protected]> wrote:
>> Right, but the auth distinction could be made along similar lines to
>> http/https - it's orthogonal. I do think at some point we end up
>> caching graphs (and their provenance to, I hope) but the bit I like
>> about danbri's Gremlin play is that it's a really stateless wander.
>> Starting from the assumption that the graph is public, we go and forth
>> as we choose.
> I don't think an implementation without caching is possible. What
> happens when you iterate through properties, an rdf:List or retrieve
> the properties of neigbouring nodes that have a hash or is bnode, I
> assume in all these cases only one request is made an then it is
> cached.
>
>>
>> You know I'm a big fan of SPARQL, but the Gremlin approach really does
>> seem to render a lot of that redundant. Ok, maybe while you're walking
>> the Web you might want to pass the data into a local SQL DB (for
>> example), but being able to walk the paths could be really useful.
> Tha's exactly the approach of the clerezza RDF DSL you use / and /- to
> wander through the nodes, adding a virtual Semweb graph (GGG) that
> dereferences resources is trivial with or without caching (but with no
> caching at all you would have sever limitations). Curious what the
> caching policy of Gremlin is.
>
>> SPARQL 1.1 has path stuff, looks good on paper, but let's say you've
>> set up your eCommerce site, being able to walk with data spectacles on
>> looks good to me.
> exactly.
>
> Reto
>
>>
>> On 12 May 2011 14:57, Reto Bachmann-Gmuer <[email protected]> wrote:
>>> HI,
>>>
>>> The .in .out seem to be equivalent to / and /- in graphnode. Now for
>>> navigating the web-of data we could simply add a virtual graph that
>>> dereferences named resources in a triple pattern adding the triples to
>>> a cache. A simple solution for authority would be the MSG, this
>>> wouldn't prevent me from saying that you know me and for this triple
>>> to be in the virtual graph as if you had asserted it but it would
>>> prevent me from linking two named resource without having authority
>>> (i.e. control resolution of the uri-space) over at least one of them.
>>> Another approach would to limit authority to the non-symmetric CBD
>>> (expanding only objects but not subjects) of the dereferenced
>>> resource.
>>>
>>> Cheers,
>>> Reto
>>>
>>> On Thu, May 12, 2011 at 2:10 PM, Danny Ayers <[email protected]> wrote:
>>>> Hi Henry,
>>>>
>>>> I did have problems seeing the relevance of your work on friendly RDF
>>>> syntax to the Clerezza project, while it's good work the tie-in isn't
>>>> obvious. But I just had a demo of Gremlin from danbri, and now I think
>>>> there's a way of pulling this stuff together.
>>>> Gremlin is a little language for graph traversal which allows you to
>>>> walk the Web of data, node by node. The key part is that as you are
>>>> going through the graph, HTTP GETs are taking place. Get that into
>>>> your Friendly code and it's a winner!
>>>>
>>>> The way I imagine it working is using the command line bits to visit
>>>> the parts of the published data from the point of view of a client - a
>>>> browser or crawler, hopefully more intelligent things too.
>>>>
>>>> Check Dan's blog post (and the addendums in comments), I think you'll like 
>>>> this:
>>>>
>>>> http://danbri.org/words/2011/05/10/675
>>>>
>>>> Cheers,
>>>> Danny.
>>>>
>>>> --
>>>> http://danny.ayers.name
>>>>
>>>
>>
>>
>>
>> --
>> http://danny.ayers.name
>>
>



-- 
http://danny.ayers.name

Reply via email to