Voynichese is a numeric cipher?

[ololololo]

or an /// chol iin chor o /// qo qo ey iin an in oke y y /// ch o /// chor an in oke y y /// cph y oke y oke y /// ch cph y o ey iin an in oke y y /// che ey in /// chor ch o in y /// ch eor cph y cph y q or an in an in oke y iin /// cph y o ey iin or in /// chol iin chor o /// chol iin ey in y /// cph y cph y ey in ey in oke y oke y oke y in /// ch ch an in y /// an in oke y /// o ey iin oke y iin /// an in oke y /// y /// chol y eor ey iin y

I decided to add an in to get rid of the ubiquitous FB, as well as to add nearby numbers (chol iin). Apparently, I understood how the algorithm works:
If YUO OOKCFBEAA to VEHA REE TOKCFDEA book VTOBA TSRRPOIHHEC RONIB YUO YMA RRMMEEEB TTHA HE ONEC HDA A WSNA

If you Facebook(?) to have REE TOKCFDEA book about Christopher Robin, you may remember that he once had a swan.

How long did it take me? A very long time! Plus, there are errors in the algorithm. But I solved the last anagrams easily, so, excluding the algorithm problems and confusing letters, it took me about 4-5 minutes.

I'd rather encrypt it manually...

---

This was the exact output from the encoder and the input is an English sentence, I didn't change any processing, I didn't add any nulls.

I've already understand 

I'm also more interested in the second example, where the extra /// marks were removed.

If you encrypt correctly and align the letters in descending order, rather than in the supernatural order as in the algorithm, there will be no problems. The word boundaries will still be "visible" because they will be where a large number follows a small number, for example. y qo, iin o, iin cph, etc.