Larry Wall: You should not take this picture to mean that linguistics is the opposite of common sense. But there are many computer scientists who have complained that it should be more like engineering and less like an art. And I disagree, but that's okay.
Enjoy life and enjoy the freedom of expressing yourself. Which means: support you as much as possible, and not force you in any way. Perl 6 tries to be even better than Perl 5 as your humble servant for translating your thoughts into executable code (if possible).
Every style and level of expertise has to be supported. And even more difficult: all of them have to play together well. The default often imitates a natural language (e.g., English). The following oft-repeated terms encapsulate strategies and concepts for this.
Perl's everlasting slogan: "There is more than one way to do it," meaning your way is valid, too. Just be aware of logical consequences.
People like to see their code as a picture at first glance. Therefore names and operators should be distinctive, similar things should look similar, different things different. Larry likes to add: weird things should also look weird. And sometimes even characters can be ideograms; like the Z meta-operator, which shows visually how it works (first left then right, then next left, etc.).
This also has to do with natural attention flow. If a topic or first word of a sentence sets an understandable context for what follows, the meaning is built up word for word. This is much easier to grasp, than if the last word of a block resolves the block's entire meaning (like regex modifiers do in Perl 5). In particular, for and given are called topicalizers, because they set $_, the current topic variable.
The same term or symbol might do different things in different contexts, although we have done our very best to consistently represent the same general idea. Maybe the best example is the * aka Whatever. But every context is its own domain, optimized to solve the problems of that domain.
Not only is syntax reused, to minimize the stuff you have to remember, but concepts are reused also, wherever possible. For example, Pairs are used to build hashes or named parameters in routines calls. There are plenty more examples. Maybe the biggest one: the regex engine that parses your code is the same engine you use in your program.
Most of the special rules that plagued Perl 5 are gone or generalized.
Perl 6 is much easier to understand, both for people and also for software parsers. The language is designed to be understood by machines in one read, dividing the source in smaller and smaller units. As a nice side effect: syntax errors can only damage the understanding of small units.
Beside TIMTOWTDI, the other ancient slogan is "Make easy things easy and hard things possible." Even if nearly every cool idea is crammed into Perl 6, we don't lose track of the basics. Features are made as accessible and free of artificial prerequisites as possible. For example, printing a variable as a line without explicit quotes and newline (say $var), or reading a file without a file handle (slurp("filename").lines). Rule of thumb: if you can think of a concept as trivial, then it should be. On the other hand, if you can think of a computable solution, it should at least be doable in Perl (no artificial walls).
This is a refinement of the last rule: common, often used things have short, straightforward names. More verbose names indicate more rare, harder to grasp concepts, or things that overwrite the default rules ("weird things should look weird"). It also implies that the defaults lean toward what a beginner (non-Perl 5 programmer) might expect ("hide the fancy stuff").
This philosophy underlies the two previous terms. All Perl words should reflect logical units which can be freely combined. Making it more complicated than necessary is obviously not desirable. But oversimplification will create artificial complexity somewhere else. Like a water bed, where you can press down in one place but the water will go up in another.
We learn, take and adapt from every source possible to offer the best of all breeds. You should be able to keep your style when switching to Perl. Perl 6 supports the structured, object oriented, aspect oriented, functional, design by contract, declarative and logic programming paradigms.
This term is an extension to the previous. Not only every programming paradigm, but also every technique or neat trick loudly marketed out there as "must have" "in the [%DATE%]", should be available in Perl 6 (or at least be possible), as long there is expressive value in it. A few examples include "named parameters", "lvalue subroutines", "parametizable subclasses" or "hygienic macros".
We are very aware that many ambitious software projects that are rebuilt from scratch fail. Especially those which aim to be "even better in all ways" successors of already successful creations. In fact, this happens so often that the term "second system syndrome" was coined. We chose this path for Perl 6 anyway, because we believe it's the only way to get a significantly better language. (Gradual improvements are flowing into Perl 5 all the time, sometimes inspired from Perl 6.) That's why the unofficial motto of Perl 6 is "Second system syndrome done right".
As a consequence of these principles, various trends can be observed.
The initial source of the Perl 6 design came from 361 RFC (request for comments) submissions by people from all over the world. Larry Wall sorted and commented on them in several documents called the Apocalypses (meaning "revelations"). Damian Conway wrote, for each Apocalypse, an Exegesis ("explanation"), that expounded Larry's ideas. These are now all historic documents that not have changed since, because they were much harder to maintain than the Synopses ("abstracts" or "comparisons" (with Perl 5)). The Synopsis documents, along with the test suite, form the official specification of Perl 6. More details on that can be found in the history section.
The Synopses are continually edited because the more we think about the syntax and semantics, the more detailed the picture gets. The other important input comes from the implementors. They uncover cases that worked only in theory, and important corner cases not thought about before. This mutual influence of specification and implementation we call the whirlpool model of software development, which is an intended pun on the well-known waterfall model.
Design changes are usually discussed in the p6l mailing list and the #perl6 channel on the freenode IRC network.