mnemonicToSeed should throw on non-string input

[dcousens]

It should probably additionally throw on empty string? But maybe someone has a use case there...

Additionally:

• enforce no duplicate whitespace
• enforce non-zero length
• enforce only printable characters

[junderw]

IMO NACK... this is not something the library should be throwing on.

The application should decide whether to throw or not on zero length / multiple whitespace etc.

[dcousens]

@junderw then let us provide an "unsafe" option.

mnemonicToSeedUnsafe

The advanced user needs protection too.

[afk11]

concept ACK - it's so easy for whitespace to be introduced, and then the mnemonic input is directly hashed.. It can be pretty easy to miss the typo 'abandon abandon abandon' and 'abandon abandon abandon' when it's a large block.

The whitespace isn't covered by the checksum, and if we allow users to make this mistake, we can expect people to panic when the recover from a seed that was transcribed to paper..

Empty string is forbidden by BIP39 - the 'recommended sizes' are the only acceptable ones it turns out, so a zero length mnemonic is invalid :)

[dcousens]

the 'recommended sizes' are the only acceptable ones it turns out, so a zero length mnemonic is invalid :)

AFAIK, validateMnemonic should make that distinction, not this function. This function should support non-strict BIP39 mnemonics, but I think the above error cases are common enough to hurt both types of user.