Re: hydra and void

On 29 October 2014 21:13, Ruben Verborgh <ruben.verborgh@ugent.be> wrote:

> Hi Melvin,
>
> > Thanks Ruben.  How about discovery.  Does hydra tell you *where* to find
> an API, from a website root?
>
> Hypermedia. Just follow a link.
>
> And if you ask me, the website *is* the API :-)
> It's time we stop treating humans and machines so differently;
> give machines the right hypermedia controls, and they can use the
> website/API.
>

Sure!


>
> > VOID does this as /.well-known/void is registered already with IANA
>
> I've never understood the need for .well-known.
> Humans don't need it, right?
> We land on a random page and just follow links to get where we want.
> The same can work for machines.
>

There's 3 ways to do discovery:

1. Follow your nose
2. Well known locations
3. Search

Just about using the right tool for the right job.  In an ideal world
everything would be 5 star linked data, and (1) would be dominant.  But we
have some work to go before that is true.

I only mentioned well known / void for practical reasons, because it
already exists and is standardized.  If there's a better way I'm open to it.


>
> > Is there any equivalent in hydra?
>
> A machine starts with a URL of some resource in the API.
> It GETs a machine-readable representation of that resource.
> Inside that resource, there will be links, which a machine can follow.
>

Got it, trying to work out how it would look.

My use case is:

<user> <predicate> <payment_processor> .

User I can do.

Predicate im not sure on.

Payment Processor I think would probably be the root domain, meaning more
follow your nose to the API.  Or I could direct to the API but there's
several so I wonder how to mark it up efficiently, I cant have a dozen
triples per user very easily.


>
> You could also be very explicit and add
>     </docs/> a ApiDocumentation.
>
> But honestly, I think a link or form works better.
>
> If you want to see such a hypermedia API at work,
> take a look at
> http://data.linkeddatafragments.org/dbpedia2014?predicate=dc%3Arights.
> This is a random resources from a hypermedia API.
> Try to get a representation both with your browser and with curl.
> You'll see that in all cases, you'll get links to reach all of the API.
>

Thanks very helpful.  Still slightly unsure what triples would work best
for discovery of *where* the APIs live, but I think I'm making progress!


>
> Best,
>
> Ruben