Okay, so putting it into my own words, to check I understand this:
the id of the webside is the root json link blob, which represents a directory.
the urls are resolved from left to right - if there is a hash, it looks it that up directly - but if there is a name (file name or directory name) it takes it from the json link blob - which probably points to another blob - if that path name is a directory, then it should point to another json-link blob - but if it's a file it should point to a content blob.

This is pretty interesting. It's an interesting pattern - you securely identify a root object, then refer to it's own links by their name. It would be interesting to also have this sort of pattern with messages. Since they also encode links and values.