Positional allography?

[oshfdk]

What's against reducing the number of letters? Especially if all hopes of simple substitution have been abandoned?

Have you ever seen Roman numerals in a manuscript? Usually, the final "I" gets a swoop added to it. It's effectively a positional allograph. Why would one argue I'm favor of it being a separate numeral, just to increase the size of the set?

Roman numerals are not a natural language. I assume Frigorifico is primarily talking about natural languages in this thread, since the original post starts with:

It is well known that some symbols only appear at the end of words, the beginning, or followed or preceded by some specific symbols. This is often used to argue that voynichese can't be a real language because no language behaves this way, but what if these aren't different symbols?

If we are talking about ciphers on the other hand, then there is nothing wrong with reducing the character set. When I was computing sequence repetition stats, I tried various alphabets, including the 2-character alphabet, where only the basic minims/curvelets encode information and all flourishes are for obfuscation.