How many words and which ones

[betatim]

This is to discuss how many words and which words to use to encode locations.

1. use 3 words, which means we need a wordlist with 2**15 or 32000 words. It seems hard to make a good one of those. At the very least such a list is full of obscure words.
2. use 3 made up words, either by markov chain or just randomly filling in CVC CVVC VCCV etc patterns. These words appear to be no more useful than those in 1
3. make a sentence like structure: noun-adjective-noun
4. use four words, means we only need 2**12 or 4096 words. Much shorter list, results in less obscure words.

[betatim]

Examples of four word encoding:

echo-filter-puts-holder mirror-autumn-quoth-peril percy-regard-empty-sugar howard-galen-genial-slavic flexed-tissue-cindy-drum georg-dead-farmer-inter cuts-cases-trucks-malawi they-yvonne-study-dialed andean-oils-wisdom-pistol dane-botany-texts-mack clarke-clone-fuzzy-galley vice-heavy-lancet-lemon gavin-gown-hanna-pull breast-hence-norton-ther accord-fangs-riding-gandhi danced-spot-sense-admit erased-sized-anthem-rating union-blogs-gaps-genre bumper-strewn-august-dixie move-rawls-barber-sword salts-slider-dance-mantra layer-george-forma-pliny

[betatim]

Examples using three word encoding:

conran-zalmay-kaylan darid-aboud-miasm devant-allier-neot dicky-andin-orgie fenris-bashes-orvis frisky-bhave-pampa heclo-carino-polos huldy-choric-recks jawing-clemen-roesch jthe-dein-seguir mbira-gopi-seibel meetup-logger-shulde meto-ludlam-skis oneal-mallik-sunne purlin-mcduff-tapirs rica-nelson-tfie rito-ochino-urbs vasu-rastus-viij wliat-shami-wlio yasha-slink-acorns anta-spiny-ayalon bernt-yoichi-barbet bilgi-aias-cenda calash-asaad-coch croson-cubano-creuse curius-daza-grise elene-ensayo-hetman figur-ericka-hiuen goleta-etsuko-jell

[betatim]

Conclsuion we go with four words per location, instead of three. Easier to remember as the words are simpler and often you only need to remember three as the first one doesn't change that much for locations close to each other (like within your city)

@kdungs ?

[kdungs]

Yes but I would really like to make it work with 3 words as well.

Can the algorithm tell us which area belongs to a location? If so, we could mark that on the map as well.

[betatim]

What exactly do you mean with "make work with three"?

decode() works with 3,4 or six. So if you go to blah.com/hello-world-green then it will take you there and the API returns the three, four and six words right now.

So for me it is more a question of what do we use by default. Thinking we should use 4 as default and display three and six as well in the info box.

[kdungs]

What I meant is the incremental thing. At this stage 3 words uses a different word list than 4 words and so on. So whether you use three, four, or six words doesn't change the accuracy just the words.

[kdungs]

So we go with four words and close this?

[betatim]

+1

Still reviewing the entries on the wordlist. See the wordlist-refining branch.

(Meta: how to reference a branch?)