This is going straight on my list: https://twitter.com/pgl/status/1405614755000295427

Now we know why it's always DNS breaking /s

Neat. I thought for a moment it was just a script embedded in a TXT record, but you do communicate over DNS with a server-side implementation of the game. The perl bit is just to unescape the response.

It uses a custom DNS server that returns the result as a TXT record for the domain <your guess>.wd.ip.wtf. You don't need to point directly at that server as it's registered as the nameserver for that domain.

Awaiting the inevitable Wordle in CoffeeScript in pure CSS on the blockchain as a Service.

I really like the idea of wordle-over-dns.

I wish all wordle authors would use words with 5 unique letters and singular nouns or infinitive verbs, though.

I see where you're coming from on 5 unique letters but have a strong preference the other way. The possibility of double letters adds a lot of interest to the game for me. But I can see how someone would find it annoying too.

I'm curious about the singular nouns or infinitive verbs suggestion though. What's the thinking there? Also to be clear, are you advocating for no adjectives/adverbs, or are you just restricting which nouns/verbs are allowed and allowing other types of word to be unrestricted?

The reason for the restriction is that you make a near endless amount of words by allowing plural, past, future, genitive (at least in Danish, that doesn't take apostrophe in the genitive) et cetera. I'm unsure how I would restrict adjectives/adverbs, but the idea is that the words should exist as main entries in a dictionary.

Here's where I come from:

Decades ago I learned to play a word game which was more like MasterMind than Wordle, and it could be played with just pen and paper (later I have played it over sms, and over gratis bank transfers):

Each participant writes down a secret 5-letter word (unique letters, singular nouns, no verbs) and take turns to guess the oppponent's word. The only clue you get for each guess, is how many correct letters your guess has. You do not get to know which letters are correct, if any, nor if they are correctly placed. It's a bit more challenging than Wordle, something that justifies the restrictions, but even with Wordle I think it's reasonable to use just the “root dictionary words”, if I may call them so.

The different forms of the word are really just a constant factor. Moreover, not all endings have the same number of letters, so it's easy to distinguish between e.g. eat vs eaten vs ate.

May I ask what is your favourite site for playing Wordle in Danish? I know someone who would enjoy that.

They are https://ordlek.github.io/ordlek/dansk/ and https://xn--wrdle-vua.dk/

The former is translated from Swedish by its author and I have promised to help the author make it look like fluent Danish, but I haven't gotten around to it just yet.

The thinking is probably that if you allow plurals and certain other verb forms to any significant degree you probably end up with a disproportionate number of trailing "S" letters--which would be less interesting.

The combination of five letters, the rules, and the number of guesses seem to calibrate the difficulty pretty well for me. Given enough time, I could probably solve most words eventually with just my brain. (In practice, I sometimes go to a dictionary program that's not wordle specific to tickle my brain once I know what I'm probably looking for.

> The thinking is probably that if you allow plurals and certain other verb forms to any significant degree you probably end up with a disproportionate number of trailing "S" letters--which would be less interesting.

There's no reason it has to be noticeably disproportionate, the clues are curated and not just picked randomly from the list of words that are legal to guess.

I guess I'm also curious if the suggestion is that all these words are illegal to guess or just not used as clues?

It appears as if there are plurals in the words you can use to guess. I just tried PILLS and it worked. I have not yet run across a plural as the answer but I haven't looked at the word list.

1.6% of possible answers end with "s" but none of those are plural. 30.5% of allowed guesses end in "s" however.

Wordle itself doesn’t use 5 unique letters though.

Some of the words I’ve picked are a bit tricky, it’s trickier than you think picking words!

I know the original Wordle also don't use 5 unique letters, that's why I wrote “all wordle authors” so you wouldn't feel singled out.

I chose words for a Danish wordle, and distilled 64k dictionary words down to around 3k.

This is really clever, you can also play in the browser with online DNS tools like https://www.whatsmydns.net/dns-lookup/txt-records?query=wd.i...

Using the example guess https://www.whatsmydns.net/dns-lookup/txt-records?query=cran...

It would be better if it just represented replies as ASCII chars to avoid perl postprocessing. Most terminals probably doesn't display colored emoji anyway.

Interestingly I just tried it on Windows. For me Windows Terminal managed emoji, but only the green square! (After running “chcp 65001”)

However the unescaping isn’t needed, nslookup on Windows just outputs the raw data. Makes me wonder if I could just use ANSI terminal escape sequences.

Here's an emoji-less version that prints correctly-placed characters in uppercase, incorrectly-placed characters in lowercase, and incorrect characters as spaces. Gotta love bitmask tricks.

wd() { dig +short txt $1.example.wd.ip.wtf | head -n1 | perl -pe's/"//g;s/\\226\\172\\156/ /g;s/\\240\\159\\159\\168/chr(255)/eg;s/\\240\\159\\159\\169/chr(223)/eg' | perl -pe'$_&=lc("'$1'")'; }

What's a good resource to gain more-indepth knowledge on DNS?

I’m not sure there’s one good resource, DNS is a complicated system, mostly as a result of evolving along with the internet over many years.

However Julia Evans has done a bunch of interesting things lately: https://jvns.ca https://twitter.com/b0rk

I'd guess Cricket's book(s) published by O'Reilly can guide you deep enough. Of course, developments since their publication are not covered, but you can still learn a lot.

DNS TTLs would make this interesting to play.