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

[Mark Knowles]

Actually it is not difficult to create unbreakable cipher: xor with random key where the key is not shorter than the cleartext does the thing.
The problem is usability. First, it’s unaffordable to create the new key every time you send new message. So you need either to use the same key every time or to generate them deterministicly.
Second, the method itself should be extremely simple, fast and reliable. From Wiki:
[Playfair cipher] was initially rejected by the British Foreign Office when it was developed because of its perceived complexity. Wheatstone offered to demonstrate that three out of four boys in a nearby school could learn to use it in 15 minutes, but the Under Secretary of the Foreign Office responded, "That is very possible, but you could never teach it to attachés."

And it appears that in practice it very difficult to create good universal cipher. And that’s why there are plenty of them

Nick Pelling introduced me to a phrase "Security through obscurity" which sounds applicable here. I have a suspicion that I could without a huge amount of difficulty produce a cipher to encode a manuscript which would be as difficult to crack as the Voynich and practical. I am not saying that it would be the same as the Voynich cipher just as difficult to decode. And I am saying that many other people could also invent such a cipher.