Historically, Rebol family takes roots in the ideas of symbolic programming, denotational semantics and distributed computing. It never was an academical project (so you won't find any papers or articles), but rather a mix of pragmatism mixed with nonacceptance of the current state of software, so there's no magic to it: you reject the bloat and keep things simple.

The core idea is that, in Red/Rebol, you don't write "code" in the usual sense of the word, but rather load a data structure in memory that interpreters/compilers then process.

This data structure is a concrete-syntax tree (CST), a product of lexing syntactically-rich data format (think of a human-centered, futuristic JSON with unique literal forms for things like IP addresses with RGBA tuples, dates in various ISO formats, monetary values, URLs, e-mails, etc). The result of evaluation, then, is the end game of interpreters/compilers "walking" this CST — and each one can interpret the same tree differently.

So, Red/Rebol is both a programming language and a data format, there is no distinction between code and data. Rebol called itself "messaging language" because this CST, coupled with semantic rules for its interpretation, constitutes an embedded DSL (aka dialect), a message that you can exchange not only between machines (as a data structure), but between machines and humans (as a readable format).

Implementation-wise, the leverage comes from the runtime library that reuses what OS provides, high degree of polymorphism, everything being a first-class citizen, and above-mentioned syntactically-rich homoiconicity with general orientation towards linguistic abstraction.

With all of that we end up with a tool (speaking of Red) that covers the whole spectrum of software development, from metal to meta: user-land functions that seamlessly operate on CST values and enable meta-programming at run-time without macros, and a set of dialects that process said CST and encapsulate the problem domain's complexity, like e.g. VID for declarative GUI specification, Draw for raster drawing, Parse for creating other DSLs (think of it as Meta-II on speeds, wearing jet-rollers) and Red/System for low-level programming. All packed into a few megabytes.

These are the principles. Rebol was an interpreted language bootstrapped from C, Red is both compiled and interpreted: runtime is written in Red/System, and bootstrapping compiler with Red/System toolchain are written in Rebol2.

So, Red is written in itself, Red/System and Rebol; Red/System is a dialect of Red; Rebol is 90% compatible with Red (roughly the same data format and language semantics). Let this meta-circularity sink in.

Balbaert's book is IMO a very shallow explanation of the language basics, often plain wrong and technically incorrect, so I cannot in good faith recommend it, especially if you are a newcomer to Rebol family with ambitions to go advanced and master the tool.

WRT pointers toward design-centered discussions around Red/Rebol, consider to join community Gitter channel and read wiki articles. E.g. one of them (shameless plug) goes a bit more into conceptual underpinnings that I described above.

https://gitter.im/red/red

https://github.com/red/red/wiki/%5BDOC%5D-How-Red-works,-a-b...