Ambiguous messages encoding

[jakubtrnka]

I went through this scheme and following crossed my mind:
Whereas for encoding BIP39 enthropy, which is multiple of 4 bytes, this technique works well and unambiguously, the left-zero-padding inside 11 bits words is problematic if we want to encode arbitrary message (we wouldn't know where starts message and where starts padding).
This problem could be solved if the 11-bit word was left-padded by all zeros but last bit which would be '1', i. e. 5bit message (xxxxx) would be padded like this 000001xxxxx. We knew we only need to discard everything up to first '1'. In case the message fitted whole 11 bit word, dummy word 00000000001 were inserted.
What do you think about this?

[iancoleman]

I'm not sure what problem you're trying to describe.

bip39 entropy is multiples of 32 bits (4 bytes), but the bip39 mnemonic (ie with checksum) is a multiple of 11 bits. Which encoding are you trying to improve?

Why does this tool want to encode an arbitrary message? It's meant for encoding bip39 mnemonics.

[jakubtrnka]

It's perfectly fine with BIP39. It was just idea so that it could be used for any secret message, not just BIP mnemonics, and were almost for free - resulting share would differ from current version with just one bit for any common BIP39 lengths. Just idea...