Refactor Tagging System

[Pritchy96]

The currently implemented tagging system is inherently flawed as discussed below. A new implementation needs to be conceived and implemented, most likely at the same time defining a set of axioms for data handling within MOIRA.

[Pritchy96]

The currently implemented tagging system is inherently flawed as a result of the Network Interface and the word processor test being developed in unison before being split; In the current system implementation, System tags are not defined by the system whatsoever, but by a client application. If a system tag were to exist, it would be defined by MOIRA itself. However, at present I see no reason for such a thing to exist.

In addition, the concept of a user tag is flawed in that it does not allow for a client set key/value pair, instead the key is set to 'userTag' by the system, that is, Moira, by default. This is incorrect! It would be perfectly fine for a client to set a tags key to whatever it so wishes, be that 'userTag', 'sysTag', or anything else, so long as it is the client which sets the tag key/value pair, and the system simply dumbly processes it, with no greater knowledge of the tag content.

When abstracting MOIRA away from the current implementation as a backend for the word processor to its true purpose (statement 1), a further issue becomes apparent; Defining the scope of a node.

MOIRA is built upon the concept that all data is available to all programs. That is, any Application layer has read/write access to all nodes (read/write access may be granted at the application layer?). However, that application layer may only expose a subset of the nodes to the program which it interfaces with: we trust application layers, because we (probably) wrote them, they're fairly simple, and do not store any data long term, unlike possible applications. We don't explicitly trust all of the applications used, especially while statement 2 is incomplete.

We also require a way in which a node can flag itself as a valid node for a given application: While there is nothing stopping an application layer serving an application with any node, valid or not, we should have some way of notifying to the application layer that this node will work with an application, be it a 'belongs to'/'origin application' tag, a parse rule field/tag, or some other method.

[Pritchy96]

One way of solving this would be to strictly define the tags in the system, such that, for example, all tags referencing a GPS value would be strictly defined as so, perhaps via its name ('position'), and the structure of such tags would be strictly defined 'a comma delimited string containing longitude, latitude', for example. This would

This would perhaps shift the structure setup from a node: a tagable piece of data, to a node : simply an ID with no intrinsic data, with tags that contain all information about the node. This way we can be much more flexible as to what a node actually is (it does not have to store something in the filestore to be valid), and also the tag system is much more rigid, as we could have dynamically defined tables which store the structure of each tag type (think analogous to MIME types at this point), which allows for our strict tag definition.

tbc