@slush0 @prusnak @voisine @ebfull

What are the chances for this to get merged soon?

Are 'Abfall' and 'Apfel' too phonetically similar? Would a blind user listening to a spoken seed be potentially confused by having both those words in the list?

@rodasmith, no, those two actually sound quite distinct (Upfall vs. Upfle would be a rough English transcription).

The words 'Graf' and 'Graph' are homophones, so one should be replaced.

Another pair of homophones: 'Miene' and 'Mine'.

Here is a few words I've thrown out when shortening to 2048 that can be put back in if more problems appear:

Abakus
Ahorn
Alpen
Antrag
Baby
Banjo
...

I'm also not sure whether 'Faser' and 'Phase' are distinct enough.

Are 'Spray' and 'Spree' homophones?

How about 'Staat' and 'Stadt'? The vowel is longer the the former but I don't know whether that is a clear enough distinction.

Spray and Spree are totally different, the latter is proununced with 'Sh'.

Staat and Stadt is a very long vs a very short vowel, hardly mistakable. Faser vs Phase might be problematic, throwing the latter out.

Are 'Uran' and 'Urahn' homophones?

No, emphasis on 2nd vs 1st syllable.

Ack (now that those pairs of homophones are removed)
I am not a native German speaker, so review by a native speaker would be welcome, but from my non-native perspective of German, this LGTM.

Is there a possibility that the user will forget that all words start with capital?

Does capitalizing the first letter have a grammatical meaning for conjugation in German?

I would say to go with whatever is in the dictionary. If nouns in the dictionary all start capitalized, then I think this is OK.

If a user has to remember case (I know some people who write in all upper case no matter what) then this is not a good idea I think.

German nouns are always capitalized, so that won't be a problem.

Abart <-> Abort
While beeing pronounced differently the only differ in a vs o.
Could be a problem when you recreate the seed from handwriting

Some more comments from reading the list (thanx for compiling it!)
"Daum" -> I'm not sure what it means. Maybe use "Daumen"
"Epoxyd" -> maybe a missspelling, should be "Epoxid"
"Faun" -> I'm not sure what it means. Maybe use "Fauna"
"Fluor" <-> "Flur" - seems and pronounces rather similar
"Kreis" <-> "Greis" - depending on local dialect pronouces rather similar
"Pose" <-> "Posse" - pronounces differently but maybe errorprone
"Spiels" -> I'm not sure what it means. Maybe use "Spiel"
"Neger" -> seems offending
"Domina", "Heroin", "Kokain" -> might be offending to some ( I imagine explaining my grandma what a seed is ab why it is important and why bitcoin is important and the seed starts with Domina Heroin Kokain ...)

Thanks @pbengert, you have some good points there, especially the offending words I overlooked. I'll try to see how I can replace them.

Forbidding words that differ by a single letter but are pronounced very differently (Abart/Abort, Pose/Posse) maybe seems a bit too restrictive to me...

@pbengert, I think I've removed all the problematic words and replaced them with innocuous ones.

@DavidMStraub Thanks for putting this list together, it feels quite good already. Since this feels like a major decision for the ecosystem and probably can't be changed easily, I want to point out a few things:

• To me the words here feel in general much more similar to others on the list than in the English BIP39 dictionary (e.g. Doge, Dogge; Wulst, Wurst, Wust; Sold, Sole, Solo; Warze, Wanze; Zelle, Cell). Technically speaking my guess would be that the Levenshtein distance between many words here is much less than in the English list, though I haven't looked deeper into that. Does someone have a tool against which BIP39 lists can be checked?
• Some of the words feel a bit hard to spell, even as a native German I wouldn't have gotten some of them right since there's multiple correct German spellings for that word (e.g. Aceton can also be Azeton, Hachse can also be Haxe depending on where in Germany you live)
• Some words have a negative connotation (e.g. Amok, Sexist, Eunuch, Geck, Geifer, Voyeur, Zicke)
• Some words feel quite "sciency" and might not be in the active vocabulary of many Germans (e.g. Spin, Binom, Axiom, Anion, Boson)
• Some outliers that have an unclear meaning or feel just very uncommon: Obrist (had to Google that), Lyzeum (outdated), Odem, Sept (unclear, short for September?)

I'll make some concrete suggestions tomorrow, feel like we could bring in some words like city names (Berlin, Kiel, Halle) or common first names instead of the mentioned ones above.

What's the reasoning behind limiting to 6 characters? Couldn't we diversify the words quite a bit if 7 or 8 characters were allowed like with the English list?

@thomasklemm, you find the words "Obrist", "Odem" and "Sept" (https://de.wikipedia.org/wiki/Sept) too uncommon, want to exclude scientific terms, but ask for a larger Levenshtein distance. And in the end we should still have 2048 words without any Umlaute or ß (this is a significant constraint). I don't think this is realistic, but in any case it would mean starting from scratch. If you are serious about this, I guess it's better to close this PR and make a new one.

By the way, I don't have the impression the English wordlist is very different qualitatively.

• arm - armed
• auto - autumn
• awake - away
• card - cart
• clock - clog
• define - defy

and so on.

To make this more quantitative, this notebook proves that the distribution of pairwise Levenshtein distances is not worse than for the English word list.

@DavidMStraub Thanks for creating this PR. I can't really contribute to the discussion unfortunately but I was just wondering are there any indications if this might get merged any time soon ?

@majuric that's a good question. At this point I find it frustrating that nothing happens even though I addressed many suggestions.

Many Thanks to @DavidMStraub for initiating this and I fully agree with his intention to achieve a BIP0039 German wordlist with low number of letters per word. One will see the advantage when engraving the mnemonic into stone, steel, etc. - it would save quite a lot of time to many people.
Actually I was not aware of this ongoing discussion here and already tried to create a list with a maximum of 5 letters per word. Pending on the underlying criteria, this wordlist was not too bad but with room for improvement.
Concerning the wordlist proposed by David I agree with @thomasklemm that there is also room for improvement, for example regarding the choice of some words. On top, there is this conflict with the Spanish wordlist which is still not clear if this is a “must-have” or “nice-to-have” criteria.
If you don’t mind, I will try to structure this topic (list of criteria) and try to come up with a revised proposal in parallel (~CW49).

Habe die Ehre,
here is a first proposal to structure the requirements. If you see anything missing or wrong, please inform.
20181219 - BIP0039 German Wordlist - SOR.pdf

As @DavidMStraub already mentioned, it’s indeed not possible to create a reasonable BIP0039 German Wordlist out of nouns with maximum 6 letters and a minimum levenshtein distance of 2 (TWO) plus all the other criteria at the same time.
There are obviously two main possibilities for a next step / evaluation:

1. Scenario 1: keep the intended maximum of six letters per word and allow some of following criteria:
   a. acceptance of more types of words than nouns (verbs, adjectives, ?)
   b. acceptance of levenshtein distance of ONE “addition / substraction” (Example: TIER & STIER)
   c. acceptance of levenshtein distance of ONE “substitution” of non-similar sounding letters (Example: ERDE & ERLE allowed / ERDE & ERBE not allowed).
   --> b & c could be limited for example to a certain amount of levenshtein collision, for example a max. of one collision (Example: MAUS & LAUS & HAUS not allowed / MAUS & LAUS allowed)
   --> or/and requiring different grammatical gender of words (Example: MAUS (feminin) & LAUS (feminin) not allowed / MAUS (feminin) & HAUS (neutrum) allowed). For verbal transmission the different article could help to identify the correct word. Example on the phone: MAUS wie die Maus in Tom & Jerry / HAUS wie das Haus in dem ich wohne
2. Scenario 2: Increase the number of accepted letters to 7 or even to 8 and allow only words with minimum levenshtein distance of TWO or more.

Other BIP0039 Wordlists solved this quite differently, most allow verb and adjectives, some allow levenshtein distance of ONE --> see attached overview.
So there are two main questions to be answered now:
A) Do you see any concerns using verbs and adjectives? If we go for capital letters of the complete word (what I highly recommend) I don’t see a roadblocker. Do you?
B) If you don’t mind, I will continue to check for a reasonable solution within Scenario 1 or do you see a minimum levenshtein distance of TWO in any case as a must have?

In order to proceed, attached a wordlist related to above mentioned Scenario 1 (keep max. 6 letters per word) based on following compromises:

• Adjectives supplementary to nouns.
• Due to high similarity of many verbs up to 6 letters to their corresponding nouns (LAUFEN & LAUF, LIEBEN & LIEBE) there’s no big advantage expected: Verbs were therefore NOT taken into consideration.
• Levensthein distance (LS) “Substitution” of 1 allowed but not two words with similar letters “B&P”, “D&T”, “G&K”, “S&Z”, “M&N” (GRAD & GRAT).
• LS “Substitution” of 1 allowed but maximum 2 collisions allowed (= max. 3 affected words).
• LS “Addition” of 1 allowed but not if double letter causing the creation of a new word (example: OFFEN & OFEN, ROBBE & ROBE, ROGGEN & ROGEN) or if an “H” is involved (RUHM & RUM).
• LS “Permutation” of 1 was possible to be excluded completely (OSLO & SOLO, FORSCH & FROSCH).

I assume it’s easier for everybody to look into a PDF-file (at least for the moment), so here’s the Proposal-2019-01-15 and the related overview of criteria (SOR):
BIP0039 German Wordlist - Proposal-2019-01-15.pdf
BIP0039 German Wordlist - SOR-2019-01-15.pdf

This is (at least close to) the best compromise between the max. length of 6 letters per word and the non-similarity of words (better than BIP0039 English WL, not so good as BIP0039 French WL) + all the other criteria. By the way: The average length of word would currently be only 5,1 letters, the English WL is 5,4 and the French WL is 6,8 --> so not bad.
@thomasklemm, @DavidMStraub and others:

• What’s your opinion on this Proposal? Would you agree to above described compromise?
• If yes, do you still see any “NO-GO-words” (embarrassing / far too unknown / too similar)? Very minor modifications might be possible, for major changes we will have to increase the accepted length per word.

Thanks @SebastianFloKa for this impressive proposal. I will have a detailed look at the wordlist and let you know in case I notice any problems (I don't expect to). I agree that scenario 1 makes a lot of sense.

On a purely aesthetical note, I don't quite understand why you want to have it all-caps. IMO that makes it much less readable (and quite ugly). I don't see any ambiguities when using the proper capitalization as there are no words that differ only by capitalization.

Good work!

So looks like we all agree that writing nouns completely in lowercase is a no-go for Germans, so this should be out of scope.

Majuscule font (= all caps, = uppercase) in a text is uncomfortable to read - I fully agree with @DavidMStraub. But for standalone information with outstanding character (importance) and need to be associated as such, it's quite common to use all-caps in German. For example the ID-card, passport and driving license of Germany and Austria: All relevant information (names, street, colour of eyes, authority, nationality, etc.) are written in uppercase. Many official documents (tax documents, opening a bank account, contracts etc.) require to be filled out in all-caps as well. This is mainly to avoid confusion.

If somebody stores his seed on an electronic device (wallet), the type of character doesn’t matter. But for people creating a physical backup of their seed, for example with beat letters (= Schlagbuchstaben in German) into steel plates or using prefabricated letters etc., it matters a lot. Allowing a mixture of lowercase AND uppercase would require a separate set of beat letters (respectively prefabricated letters) and cause a more confusing and time consuming procedure to manually create the Backup. As the intention of a BIP0039 Wordlist is to provide an easy to handle solution to users, the amount of potential characters should be kept as low as possible - which would be fulfilled by using all-caps only.

Then the issue with same words as noun or adjective or verb, and there are quite some of them: Husten/husten, Klasse/klasse, Nutzen/nutzen, Rennen/rennen, Wissen/wissen etc.. There will always be uncertainties about the correct writing if a mixture between lowercase and uppercase is allowed, particularly if words are transmitted verbally. If only uppercase characters are set, it doesn’t matter which meaning somebody connects to a word (such as homonyms), the orthography will always be predefined and therefore unambiguous.

If restoring of a partly unreadable seed is necessary, it could be helpful as well. If there will exist many different wordlists in the future (a tendency we can already see today), it would help to distinguish the BIP0039 German Wordlist from others. Software tools (wallets) could take advantage of this difference and provide more relevant word proposals to the user and request for uppercase letters (or exchange automatically) if German is pre-set as language.
To be fair, the last point is an advantage (or even neutral point) and not an argument to go for majuscule (uppercase) for the BIP0039 German Wordlist – but IMO the others are.

Thinking about pros and cons of limiting words to 6 letters it might make sense to extend this to at least 7 or maybe even 8 letter per word. I think it's worth it.
I could work on a new proposal around August this year...

What's the current state of this issue? I would offer my help

As far as I'm concerned, I've abandoned this. It's really a weird PR. First, I get a few useful comments, make a few improvements. Then, I get a list of requests from @thomasklemm, some of which I proved even with a Jupyter notebook are not realistic/reasonable. No reply. Half a year later, @SebastianFloKa comes with a totally different suggestion, but does not actually submit a PR.

@FaustmannChr, thanks for offering your help, but if I were you I wouldn't invest time unless one of the maintainers confirms that this actually has a chance of getting merged.

Yeah @DavidMStraub, I think you are right...

And many thanks for the summary and the fast response!