Date: Fri, 5 Jan 2018 18:08:50 +0000
From: nullius <nullius@nym.zone>
To: Sjors Provoost <sjors@sprovoost.nl>,
	Bitcoin Protocol Discussion <bitcoin-dev@lists.linuxfoundation.org>
Message-ID: <55374d9210feebf2f758d4e7b21849ee@nym.zone>
References: <57f5fcd8644c6f6472cd6a91144a6152@nym.zone>
	<BB3FA46E-AA09-4A60-9D0F-8E350015E107@sprovoost.nl>
MIME-Version: 1.0
Content-Type: multipart/signed; micalg=pgp-sha512;
	protocol="application/pgp-signature"; boundary="s2gulaj2zoqb3hzc"
Content-Disposition: inline
In-Reply-To: <BB3FA46E-AA09-4A60-9D0F-8E350015E107@sprovoost.nl>
Cc: arachnid@notdot.net
Subject: Re: [bitcoin-dev] BIP 39: Add language identifier strings for
 wordlists
Precedence: list


--s2gulaj2zoqb3hzc
Content-Type: text/plain; charset=utf-8; format=flowed
Content-Disposition: inline
Content-Transfer-Encoding: quoted-printable

On 2018-01-05 at 16:04:10 +0000, Sjors Provoost <sjors@sprovoost.nl>=20
wrote:
>I=E2=80=99m not a fan of language specific word lists within the current B=
IP-39=20
>standard. Very few wallets support anything other than English, which=20
>can lead to vendor lock-in and long term loss of funds if a rare=20
>non-English wallet disappears.
>
>However, because people can memorize things better in their native=20
>tongue, supporting multiple languages seems quite useful.
>
>I would prefer a new standard [...] A replacement for BIP-39 [...]
>
>[snip]
>For my above point and some related ideas, see:=20
>https://github.com/satoshilabs/slips/issues/103

You present some interesting ideas; and I will be much interested in the=20
Github issue you referenced=E2=80=94thanks for that.  However, this discuss=
ion=20
is *far* beyond the scope of my simple proposal and request to add=20
standardized native language and short-ASCII identifier strings to the=20
BIP repository.  I suggest that readers solely interested in the=20
existing BIP 39 standard and its direct application to Bitcoin should=20
stop reading right here.

----

That being said, I should briefly address some of the issues you raise=20
(with further discussion best continued elsewhere):

I *strongly* urge the importance of language-specific standardized=20
wordlists.  Even an individual who has secondarily acquired reasonable=20
fluency in English will most likely have the least difficulty=20
memorizing, transcribing, and otherwise handling a =E2=80=9Cmother-tongue=
=E2=80=9D=20
mnemonic.  Such an advantage is important in applications whereby even=20
slight errors can be fatal, and every bit counts.  This is to say=20
nothing of persons who have limited or no English-language knowledge.

Yet for multiple reasons, multilanguage support is only feasible with=20
standardization.  Wordlist creation is a highly specialized task. =20
Independent implementation of standards is imperative for avoiding=20
implementation lock-in; and independent implementors (such as I) would=20
be unable to create sets multi-language wordlists on their own, anyway. =20
For a view of the language-specific process involved in creating a=20
wordlist, I invite everybody following this discussion to review BIP=20
repository PRs #92, #130 (Japanese); #100 (Spanish); #114 (Chinese, both=20
variants); #152 (French); #306 (Italian); #544 (Korean, rejected); and=20
#570 (Korean).  The rejection of #544 for Korean, and its superseding=20
with #570 is particularly instructive.

With standardized wordlists, independent implementation is easy.  In my=20
own implementation, the language switching backend (excluding the UI[1])=20
for multilingual mnemonic generation required only relatively small C=20
code changes, as seen here[0]:

[0] https://github.com/nym-zone/easyseed/commit/5b6a6668458d96d6ccc4bf19e4f=
d40fe6ea72fec#diff-20dcf1782b7568b85ea01ed695abeb02

[1] https://github.com/nym-zone/easyseed/commit/1a6e48bbdac9366d9d5d1912dc0=
62dfc3f0db2c6#diff-20dcf1782b7568b85ea01ed695abeb02

Admittedly, the multilingual requirements for seed generation will take=20
a bit more; and my nonstandard, non-BIP39 ideas which require decoding=20
words directly back to bits will take yet more.  But it is still not=20
problematic.

I only began writing this tool one week ago, as of today; and it has=20
been a side project requiring small amounts of time, not a full-time=20
dedicated task.  When I fully complete all aspects of seed generation,=20
then users will have the option of another simple open-source tool which=20
*will* be able to output a binary or BIP-32-formatted (=E2=80=9Cxprv=E2=80=
=9D) 512-bit=20
seed, given input of an existing mnemonic in any language supported by=20
official BIP 39 wordlists.  Output can then be imported to any wallet=20
software which supports BIP 32, regardless of the wallet=E2=80=99s langauge=
=20
support (and whether or not the wallet supports BIP 39 at all).

**The ease of creating such tools squarely answers your concerns about=20
vendor lock-in.**  And yes, it=E2=80=99s easy.  I can attest as a lone code=
r,=20
it=E2=80=99s easy for me to create =E2=80=9Ceasyseed=E2=80=9D as a side pro=
ject!

Finally, aside:  In the discussion at SLIP repository issue #103, I see=20
mention of m-of-n SSSS.  I have been mentally whiteboarding just such an=20
application involving mnemonics.  Watch for it.  <g>  It is likely that=20
I will crib the BIP 39 wordlists, given the impossibility that I could=20
create my own set of wordlists in many languages.  I only wish that the=20
BIP repository had support for more languages.  More!  Adding each new=20
language to my implementation(s) will require approximately one-line=20
code changes for me.

(Aside further:  Why is there not a Dutch wordlist?  I should like to=20
add that, please=E2=80=94meneer Provoost.  More wordlists!)

Aside still yet further:  Should you be interested in more general=20
applications of mnemonic phrases for pseudorandom strings, I think you=20
will like this future feature which currently exists only as an Easter=20
egg, (un)documented in my commit log:

https://github.com/nym-zone/easyseed/commit/ba77be1b1a1f0c6af50ceba5c89f4ad=
ece7e5dff

Further discussion is invited by private mail, in an appropriate public=20
venue, or otherwise not on a bitcoin-dev thread which makes a simple=20
request and proposal as to the existing BIP 39 standard=E2=80=94thanks.

--=20
nullius@nym.zone | PGP ECC: 0xC2E91CD74A4C57A105F6C21B5A00591B2F307E0C
Bitcoin: bc1qcash96s5jqppzsp8hy8swkggf7f6agex98an7h | (Segwit nested:
3NULL3ZCUXr7RDLxXeLPDMZDZYxuaYkCnG)  (PGP RSA: 0x36EBB4AB699A10EE)
=E2=80=9C=E2=80=98If you=E2=80=99re not doing anything wrong, you have noth=
ing to hide.=E2=80=99
No!  Because I do nothing wrong, I have nothing to show.=E2=80=9D =E2=80=94=
 nullius

--s2gulaj2zoqb3hzc
Content-Type: application/pgp-signature; name="signature.asc"

-----BEGIN PGP SIGNATURE-----

iHUEARYKAB0WIQSNOMR84IlYpr/EF5vEJ5MVn575SQUCWk+/LwAKCRDEJ5MVn575
SXfHAQDhSEq1659zkl8leF3zg8VJ3lnXA6KxE0N6o18SZDNvyQD+M0RXjCpb/VNp
vnOtlVGqXhg0PM1Jg3NBDloZX8H7Qwk=
=yhxj
-----END PGP SIGNATURE-----

--s2gulaj2zoqb3hzc--