GitLab

Primarily, being commital and direct about how special values, erm...
*aren't* -- namely, "/" is a valid segment, and so is empty string --
and confessing how much tricky work that makes.

Several methods are now more upfront about how much they *do not do
good things* on such values.

Added some new constructors for Path which *do* work for all possible
values.  (You could've handled such tricky values with a series of
Join calls, previously, but that would be ergonomically grueling.)

These facts are all defacto 'true' to the best of my knowledge now.
However, the IPLD specs repo is a bit on the quiet side about them
at present (precisely because it's such an irritating little nest of
edge cases and fun stuff)... and so the comments are also littered
with "this may change" warnings.  Still, it's better to be accurate
about what the code does and does not do in its current state.

I'd *like* to formally specify an escaping system and canonical
string encoding for paths.  That should start in the IPLD Specs
repo, though, and involve fixtures, so I won't start it here now.

Thanks to @ribasushi for the kick in the shins that these docs needed
work and clarifications.