A semantic encoding system?

[Jorge_Stolfi]

My well-meant suggestion is to first of all start wondering if the character groups separated by spaces in the Voynich MS actually represent words.

As I remember, the main argument is that the labels are often just one "word", which has the same structure as the "words" in the text.

Another (weaker) argument is that the number of distinct "words" is compatible with that theory, and so is their Zipf-like frequency distribution, and hence the entropy per "word". 

And of course I have my own variant of that theory, and my arguments for it.  But neither seem to be accepted here...

All the best, --stolfi