Thoughts on tagging design

Brian’s comment is that representing non-valued tags (referred to as “marker” tags, similar to “marker interfaces” in Java), are really more like OBIX Contracts than sub-objects. Tridium agrees with him, that we should probably represent marker tags as Contracts.

· A primary reason for this is that marker tags generally represent the “is a” relationship. For example, the “point” tag applied to an <obj> means that this object is a “point”, and conforms to the semantics associated with that type. This is the way that Haystack uses marker tags, and while that is not the motivating factor, it does capture how people would likely intend to use tags, and build on top of OBIX. A consumer of an OBIX object would like to look at the ‘is’ attribute, and quickly know what things it can do with the object. Similarly, it makes things much easier for the publisher of an OBIX Contract and for a server producing an OBIX object conforming to that Contract to be able to describe it using tags for models like Haystack, and Contracts when presenting to OBIX.

· Note that the use of the colon in contract names is a formal XML namespace usage (see the OBIX Encodings document Section 2.6), so concerns about ‘almost-namespace usage’ do not apply.

· Another reason, clear from his comment example, is brevity. Although there are more compact encodings than straight XML, the bulk of OBIX communication is today carried out using XML, and will likely be done so for the near-term future.

Would like to discuss this on Thursday if possible.

Thanks

Craig

obix message