Any chance for nomenclator?

[Anton]

Also the apparent intra-vord structure would need some kind of explanation

The intra-vord structure may serve the purposes of locating vords in a large nomenclator. If one produces the plaintext-to-Voynich dictionary (which serves the encoding purpose) as he encodes the plain text, and uses let's say the alphabetic principle of sorting the plaintext words in that dictionary, then he also must provide for some method which would allow for convenient decoding of the ciphertext afterwards - that is, a method of quick lookup of vords in that (huge) dictionary.

When we translate e.g. between English and German, we use two dictionaries: the English-to-German dictionary and the German-to-English dictionary. In both of them words of the source language are sorted alphabetically and thus can be found quickly. In the hypothetical Voynich codebook only one of the sets could have been sorted alphabetically (or in any other convenient way) - either the underlay vocabulary or the Voynichese, and I suspect that it would rather be the underlay language vocabulary, because when you encode every possible word, then you don't know in advance which words you will need, and then you are bound to composing your underlay-to-overlay dictionary just as you write the text, not prior to that. Then for the reverse task of decoding (when you decide to read the ciphertext that you've written), how would you approach it without some means of quickly finding each Voynichese word in your codebook?