With #perl6's help (in particular from putter++ and TimToady++), I have converted the layers from yesterday's bullet points into this diagram (click to zoom in -- thanks to Gordon McCreight for the thumbnail).
The Perl 5 Sugar layer is really key: it ensures that CPAN authors can use the relatively stable Perl 6 semantics without worrying about the more experimental Perl 6 syntax, and also ensures that v6.pm generates idiomatic Perl 5 code as output.
Each layer depends only on the layers below it. Everything above the XS layer will be pure-perl and hence very portable. The XS layer currently depends on some underdocumented areas of perl core (PL_check, for example), and we'd like to document them and bring them into the core regression test suite, so that they become fully supported in the future.
The diagram currently only lists existing modules, so there are a lot of "..." to be filled in each layer -- in particular we need sugar for all of semantic-layer modules -- but that's an exercise for tomorrow. :-)