What is the technical definition of theoretical computer science? What subfields are included?

Question

What is the technical definition of theoretical computer science? (Or, what should it be?)

What main subfields does it include, and what is the commonality that separates them from the rest of computer science?

More specifically: if some particular research has direct practical motivations, goals and outcomes but mostly involves very abstract methods, is it theoretical computer science or not?

Two examples to consider:

"Dual quaternions for rigid transformation blending" (Better mathematical representation of rotation and transform for animation) https://www.cs.tcd.ie/publications/tech-reports/reports.06/TCD-CS-2006-46.pdf

"Relational Semantics for Effect-Based Program Transformations with Dynamic Allocation" (Complier optimisation via denotational semantics): http://research.microsoft.com/pubs/67977/ppdprelational.pdf

[The Wikipedia article gives only a vague definition and a long list of subfields. Should just accept that there's no better definition than this? http://en.wikipedia.org/wiki/Theoretical_computer_science ]

EDIT: I guess this question comes down to "What does the term 'theory' mean in the context of computer science?". Looking at the 6 different meanings of the word at wiktionary, I don't think any of them fully fits. I guess the mathematical sense of a theory fits well for completely mathematical fields but not for others, and for VLSI, machine learning and computational biology from wikipedia:TCS it basically doesn't fit.

Answer 1

This link contains a list of subfields: http://arxiv.org/corr/home, I won't reproduce them here as the link may change, and it would be redundant.

Also, I'm reminded of the quote of someone, can't remember who, along the lines of:

Mathematics is whatever mathematicians do

It would seem to apply.