bsimons14
commented
4 years ago Thanks for the discussion Dave. Yes the boundaries between 2 and 3 are blurred, if they exist at all. I'm not convinced however that "we would expect most if not all of tier 2's meta-information to be contained in tier 3." Looking at this from an 'old-school' GML view I think the analogy is that tier 2 is like returning a GetFeature request. That may contain everything we have about the feature in-line, or it could contain only a minimum of data (e.g. name, description, identifier) and every other property (tier 3) byReference. Again this deferred response could contain everything or a minimum with everything else byReference. (Note that it need not, and probably won't, contain all the tier 2 information.) GML communities have specified 'profiles' in attempting to reach a balance between overwhelming the response by trying to deliver everything and overwhelming the search by making everything via links. So I think what we are trying to do with tier 2 in SELFIE is provide a (community agreed?) first IR response for a NIR, with full deferred resolution, but where the resolution is not just one resolution 'type' (in my example GML) but any type or representation or profile or ... from any provider. If so, then yes tier 2 is a 'convenience layer'. It then requires mechanisms and standard concepts to allow identifying what the user/crawler can expect if the links are followed when known (cf GML property tags in xlink:hrefs). Or if not known, here's a link to something about this NIR, but I don't know what you will find there (="subjectOf"?).
dblodgett-usgs
rob-metalinkage
Conceptually, the "non-information" "meta" and "data" "resource" scheme discussed in SELFIE is useful, but the language is wrong when applied to the terminology of particular technologies (HTTP Resources). As such, we need to change our language to more accurately reflect what we have found to be useful ways to deal with this three-tier model.
Everything in this scheme is identified by a HTTP URI. We have:
1 is clear -- there should be URIs that only ever return a 300 series redirect (or not, we've noted the conceptual and tactical drawbacks to that practice elsewhere).
2 is quite squishy. It is defined by the context of retrieving it rather than by specific characteristics of its content.
3 is clear in most cases but has potential overlap with 2 -- this is the focus of this issue.
Self describing data MUST contain metadata. Given that, done right, the data tier is the web of self-describing data and we would expect most if not all of tier 2 to be contained in tier 3.
If we think about it that way then tier 2 is a convenience layer. Thinking about it this way -- it's convenient for what? The answer to that seems to be crawlers and humans looking to get an idea of what a real-world feature is, what it's related to, and if there's anything interesting available for it.
Resource is the wrong word. It's ok for tier 1, but it breaks down for tier 2 and 3. A single URL may have one or more meta resource-representations and one or more data resource-representations -- each intended for a different use pertaining to the same real-world feature. This is not saying that tier 2 and tier 3 are always to be represented variously based on the same URL -- they very well may be represented as different resources (this technical diversity is what we want to enable?).
At the end of the day, what has been referred to as the "three tiered SELFIE resource model" is actually a sort of conceptual content model that we are implementing with technology du-jour.
Can we call it a content model? What's a better description for this break down?
Can we describe how the content model can be implemented in a wide variety of ways while illustrating the few key places the implementations share common practices?
What are those common practices?
How do we illustrate that this approach can bring interoperability in key use cases to a diversity of data systems with requisite diversity of implementation patterns?