How easy is it to create a cipher which is very hard to break?

[Mark Knowles]

...

I would have thought that there are lots of relatively simple methods by combining the use of filler text with another simple rule to produce Voynich like text where the morphological regularities observed in Voynichese words are much more a function of the filler than the text.

The VMS text may seem deceptively simple, but try to create a rule-set that mimics the properties of VMS text as closely as possible. If you feel it's mostly filler-text, then the rule-set should mimic the filler text rather than any small proportion of meaningful text that you think might be hidden within it.

It doesn't require a formally defined rule set. I think a lot of it could be done on an ad hoc basis. This could be a situation where I know what the meaningful text is and then fill in something made up, but conforming to the format of Voynichese words. There would very likely be multiple possibilities for the filler text, rather than unique possibilities pre-determined by an algorithm. In so far as it would be formalisable the filler text would be generated by a combination of conformity to the Voynichese word structure and a very large random element of choice. However I don't think some random organised procedure for generating a filler text like a cardan grille is required just human invented nonsense in the style of Voynichese words would serve the purpose.

Of course there are myriad ways in which filler text could be implemented, so I am talking about a general principle than a specific example.