Re: adding hypermedia to spatial data best practices

I hope we don’t lose sight in the discussion of REST and “webbiness” that most of the present Web only works because there are humans reading web pages and interpreting link semantics from the textual / visual context of both the link and the target. A link that a machine follows from one collection of data elements to another collection of data elements in performance of a task has a significantly different set of requirements. It would be lovely to have a universal hypermedia data format, for example, but so far the only technology that has come close has been RDF, and that has definitely been an uneven and incomplete success.

One of the nice aspects of a RESTful API approach is indeed the option for a person to be able to browse holdings / capabilities to get a sense of the content or how best to write clients. It doesn’t mean that same hypermedia approach is really effective for machines to use. 

I agree that some mixed human / machine use cases might capture actual usage patterns more effectively.

Josh

> On Aug 2, 2015, at 7:41 PM, Rob Atkinson <rob@metalinkage.com.au> wrote:
> 
> 
> Yes - am thinking mainly of the "deep dark data web" - but links to and from hypermedia resources should be handled consistently - so its still probably worth starting with the (not clear best practice :-) ) mechanisms to make link semantics work in machine-traversable data, then seeing what happens when including hypermedia resources into the fray.  A use case identifying an access and  processing scenario that uses both data held behind APIs and as linked hypermedia resources might be useful. 
> 
> making sure loose, but easy to use human interpratable information can be included is fairly easy, since there thats already there and we'd be looking for a minimal set of practices to layer on top of it
> 
> 
> On Sat, 1 Aug 2015 at 10:41 Erik Wilde <dret@berkeley.edu <mailto:dret@berkeley.edu>> wrote:
> hello rob.
> 
> one thing may be good to keep in mind. you say:
> 
> On 2015-07-28 22:38 , Rob Atkinson wrote:
> > Personally, I think the relationship between "data" and "hyperlinking"
> > needs some greater care.  In a self-contained database, relationships
> > are a first-class concern - however there is a prevalence in the linked
> > data world of using ad-hoc approaches to generating hyperlinks - for
> > example using owl:sameAs to link to an interactive mapping application
> > via geographical coordinates. using very general link semantics
> > "rdf:seeAlso" for links to related data is another common pattern. The
> > lack of a demonstrably good practice is fairly hard to reconcile with
> > any potential to be able to use such links in any automated fashion, so
> > the development of best practice discussion and exemplar resources is an
> > important step to take. fortunately, the Linked Data web is still tiny
> > compared to the problem space, so there is not a huge amount invested in
> > sub-optimal approaches.
> 
> it seems that most comments come from a linked data perspective. that's
> fine, but as you say, that's a tiny percentage of what's currently going
> on on the web, and what's probably going to be on the web for the
> foreseeable future. and there's an interesting difference:
> 
> - in linked data, formal link semantics (and tightly defined semantics)
> are a least a goal, even though they may not always be as well-defined
> as one might want.
> 
> - on the web, link semantics are purposefully very loose, so that link
> relationship types can be reused. what matters then is how those fuzzy
> semantics are used *in the context of a given service*. this is very
> different (and intentionally so) from the linked data goal of defining
> and then using context-free meaning that can be used globally outside of
> a well-defined scope.
> 
> now, whether one prefers the first or the second approach of "defining
> link semantics" is an interesting question in itself (both with
> interesting challenges and implications), but not something i want to
> discuss here. but it would be great to see that if something claimed to
> be best practices for the web (and not just RDF-based models), that it
> would be inclusive in terms of embracing both approaches.
> 
> > I think a "star" that matters is missing - which is to make the meaning
> > of hyperlinks explicit and discoverable - this is far more useful than
> > putting the data into RDF per se, but one could argue thats the
> > underlying intent of using RDF, in that such links have URIs for link
> > predicates - and there is an implication regarding what those URIs
> > should resolve to.  Maybe there is some good practice out there
> > somewhere of how to hyperlink without losing information or adding more
> > noise to the system we could point to - but I haven't seen one in the
> > geospatial domain.
> 
> if you just talk about the web, then all that web architecture requires
> is to make links explicit, and link relation types as well, so that
> those links are typed and can be traversed by clients. i guess i should
> add "use (explicitly or implicitly) typed links" to star number four in
> the web data principles...
> 
> https://github.com/dret/webdata <https://github.com/dret/webdata>
> 
> cheers,
> 
> dret.
> 
> --
> erik wilde | mailto:dret@berkeley.edu <mailto:dret@berkeley.edu>  -  tel:+1-510-2061079 |
>             | UC Berkeley  -  School of Information (ISchool) |
>             | http://dret.net/netdret <http://dret.net/netdret> http://twitter.com/dret <http://twitter.com/dret> |