At one end of the spectrum all things GETable are URIs without a query part. Imagine for example all of your images having an URI, easy. Now imagine an URI for all the possible sizes of all those images ever requested by an URI user. That's a lot of URIs to keep track of!
At the other end the URIs for representations are reflections the
representation. image.png?width=128&height=128&aspect=original for example. This looks pretty familiar and the representation can be computed from the URI, fewer URI's to remember.
Naming things is tiresome and once you name it, you need to remember it since someone else might use it. Mint a URI for a thumbnail of an image that you plan on using once, and you need to keep track of that URI forever, just in case something saw you use it.
Queries (like POSTing of SPARQL) are good because you describe the properties the thing has you're talking about, no need to name. And queries can be complex, like a thought.
Problem is URIs are handy (you can <a href=" to them), and maybe you can't always generate a query (you can only GET for some reason (caching?)).
URIs are like memory addresses of the great machine that is the Web. Like an application bound to a single OS, memory addresses are expected to work all the time, and the same should be true for the Web. Difference is, OS memory gets released occasionally so addresses can be reassigned. You don't know though how long an application on the web might be used (like an URI written in a love letter way back in 1995), so addresses on the Web must/should be indefinate.
That's a lot of addresses! But that's what we want, isn't it? For the Web to answer any question we throw at it? Our systems then have to maintain that knowledge.
Seems we don't hold queries to the same standards as URIs. If we ask a URI to serve up a representation, we expect to get one, and if we don't, it's broken. If we don't get an answer from a query, it's okay; the system doesn't have an answer to our query.
Maybe this can stear how we use the seamingly interchangeable addressing techniques? If it's something we'll rely on to exist in the future, like an image in a webpage, it gets a uri. If it's something more variable, give me an image that hasn't been anotated (an they've all been anotated), it won't have a URI and we'll rely on a query?
I don't like the system having to try an make such judgement; we just can't have that kind of perspective.
I'm not sure what this means.
Loose coupling, good.