Dear Professor Garfinkle,
I completely understand seeing a similarity between this approach and causal sets, but at the level of implementation and exact detail they are quite different. So, for example, in the causal set approach there is essentially zero chance for any pair of events to be null related, and in this approach to a discrete structure all the fundamental space-time geometry is null. Also, as you note, in the causal set approach the idea is to get the conformal structure from the causal graph and then fill out the rest of the geometry by a volume measure. But in this approach to a discrete Relativistic space-time (which is not described here) one does not add a volume measure but rather a measure corresponding to the Interval measure on continuous paths. In a word, the basic metrical notion is length rather than volume.
Let me answer your questions.
Well, actually what I have said above answers 1) for the discrete case. The discrete case is nice because (as Riemann observed) the discrete case comes equipped with a measure: counting measure. But you have to be careful about what you count! In this case, one does not count nodes (which is what the causal set people do to recover a volume measure), and what you are quantifying is not volume but length. This may be a bit cryptic, but in short in the discrete case one can define a "corner" in a continuous path through the space-time as opposed to an "unbent segment", and when quantifying the length of a path one counts corners rather than nodes in the path. The result is that the number of corners on a path that lies on the light cone is zero (even though the number of nodes may be unboundedly large). The connection to the Interval should be obvious.
In a continuum, the measure must be imported from outside (just as Riemann said). I have not tried yet to somehow analyze that measure in terms of anything else in the continuum case: it is just an additional piece of space-time geometry supplementing the conformal structure, just as in GTR.
To the second question:
Investing a space with a Linear Structure automatically invests it with a standard topology, via the definitions I give in the appendix. So it is easy to tell if you have a manifold. But in fact, even familiar continuous space-times (e.g. Minkowski) turn out not to be manifolds once one takes account of the directionality of time. One of my points is that the whole idea of a manifold arose in the context of purely spatial (Riemannian) geometry, and the use of those mathematical tools to deal with space-time (Lorentzian) structures is a mistake. So I am not even aiming at recovering manifolds: I am aiming to recover geometries with an intrinsic light-cone structure, which a manifold does not have.
To the third:
One of the main advantages of my approach is that one can deal with contiinua and discrete spaces using the very same analytical tools and definitions. No one knows whether at (say) Planck scale space-time is discrete or not. So it would be nice to articulate theories in a mathematical form that can be adapted to each possibility. As things are, one uses manifolds and differential geometry for continua and graph theory for discrete spaces. Graph theory (including infinite graphs) is a special case of the Theory of Linear Structures, but the Theory of Linear Structures can be used to analyze continua as well. It also allows you to take ideas developed in the context of continua and see how they play out in a discrete setting. So I think that is one thing that is an advantage.
One other advantage, which I do touch on, is that the Theory of Linear Structures allows for a distinction between intrinsically directed and intrinsically undirected geometries, which is relevant to the description of time. Indeed (as I mention) it even allows one to make a distinction between intrinsically directed and intrinsically undirected topologies, which (to my knowledge) no one working in standard topology has made, and maybe cannot even be drawn without using the resources of the Theory of Linear Structures. So if space-time is discrete (which it may be) and if time is intrinsically directional (which I think is obvious!) then the analytical tools available are better than standard topology and hence manifolds.
Cheers,
Tim Maudlin