[mvdtnz]

The user's input didn't seem to have any relationship with the text output? Obviously there's no simple one to one mapping but it would have been much more fun to use the actual letters from the phone with some kind of software to guess the intended word.

[danbruc]

He is dialing the ASCII codes, 114 for r, 108 for l, 100 for d, and 33 for !. Or maybe Unicode code points, impossible to tell from this demo.

[aschmelyun]

You're correct with ASCII! My original idea was to use T9, but my ultimate goal is to use this as a way of writing code.

Currently waiting on a small oled screen to print out the # inputs and corresponding last-typed character.

[m463]

I think on a PC keyboard you can simulate this by holding ALT and typing 114 on the numeric keypad

I wonder if you can change the rotary phone to octal? that would be great to boot a pdp 11/34.

[jstanley]

I'm guessing it works out how many digits to expect for each codepoint by looking at the first digit (i.e. if it starts with a 1 then it has to be 3 digits, otherwise it is only 2).

Does this mean it's impossible to type a newline character, or a backspace, for example?

[danbruc]

You could dial 013 and 010.

[jstanley]

Ah, so the logic is that if it starts with either a 0 or a 1 then you need to type 3 digits, otherwise 2? Nice.

[danbruc]

No idea, I didn't build the thing, just suggesting how it could work. If you want to cover [extended] ASCII, then every first digit could indicate a one or two digit code, one and two also a three digit code. So one possibility would be to have three digits for first digits zero, one, and two and otherwise two digits. All single digit ASCII codes require two leading zeros and all two digits ASCII codes starting with one or two require one leading zero. You could of course also make everything three digits with leading zeros as needed, would be easier at the cost of more dialing. Or use a timeout, if no new digits arrives within a specific time, then a new character starts.

[aschmelyun]

You both are 100% correct! Originally I was going to have a wait (say, no new digits after 3 seconds, mark the ascii character), but I figured just scrapping everything after 20 as an immediate character and anything under as a 3 digit one, it made typing a lot faster.

[onecommentman]

A few decades ago, you could make ASR-33 teletype printer with an acoustic modem print…something…by whistling into the modem. Why go mechanical when you can be artisanal/organic?

[SanjayMehta]

T9 text with a rotary phone. That would have been fun.

https://en.wikipedia.org/wiki/T9_(predictive_text)

[xxs]

I thought it'd be T9 - that was a glorious system.

[dontlaugh]

It was useless if your language isn't supported, though. For many, T9 was no longer relevant by the time their language got support.

[aschmelyun]

I still miss it to this day, for basic texting it was the superior system.

[Symbiote]

It's easy to install a custom keyboard on Android, at least.

[xxs]

no real tactile feedback

[xxs]

>there's no simple one to one mapping

Why not: you can count the letters A=01, Z=26 and use that, use 2 digits each time, numbers start w/ 3 (e.g. 0=30, 9=39)

That's inefficient but trivial to understand, then there is the T9 which use to be the very standard for the first mobile phones. It's language dependent obviously for prediction but pretty cool, too.

Then there is Morse code which would be implement as well. Overall I'd not call the 'demo' impressive at any form.

[dghlsakjg]

He’s using ascii codes.

Overall, I’d not call your “analysis” impressive in any form.