From the abstract:
This thesis presents a type system combining ML-style parametric polymorphism and subtyping, with type inference, principal types, and decidable type subsumption. Type inference is based on biunification, an analogue of unification that works with subtyping constraints.
Basically, the author has found a way to infer subtypes without using constraints, and instead via algebraic construction. Pretty interesting, tho I haven’t worked my whole way through it just yet.
http://lambda-the-ultimate.org/node/5393 has links to a more summarized paper with the basics. excited to work my was though the thesis as the paper assumes some knowledge that I’ve found it impossible to google.
You should consider asking here for references if that can help. I guess that most questions would have answers either in textbook introductions to type systems (for example Benjamin Pierce’s Types and Programming Languages, or Didier Rémy’s lecture notes), or in previous work on the topic (for example François Pottier’s thesis). That said, looking at Stephen Dolan’s thesis is indeed a good idea as wel, as theses, being liberated from page number restrictions, are often more accessible than condensed papers.