Hacker News Reader

Comment by jvanderbot | Hacker News Reader

My wife, god bless her, is really good with people. I try very hard to mimic her, but this multi-round interaction has completely permeated how she interacts over tech, even with tech. I cannot manage it.

Her: "Alexa, add to shopping list". "OK, what should I add for you". "Peanut butter". "OK, peanut butter added, what else?". <long pause while the house has to be quiet until alexa times out>.

Me: "Alexa add peanut butter to shopping list". "Peanut butter added".

Some people are TCP. Some are UDP.

View on HN

“Alexa add peanut butter to shopping list" <silence>

“Alexa add peanut butter to shopping list". “Peanut butter added. Peanut butter added. Peanut butter added.”

This is what you get when you don't check for PIPELINING in the EHLO response like you are supposed to. (-:

"Alexa add peanut butter to shopping list".

"It's currently raining would you like to see the forecast for tomorrow also I found this routine you might like would you like me to enable it"

writes peanut butter on a piece of paper

"Some people are TCP. Some are UDP."

Nailed it! This is going up on the wall in my office.

> are UDP. some

Misses, sure. But it should be quite rare that the order is wrong. It would only happen if the route changes.

Eh, comedic posts should probably be allowed some leeway in their attempt at being funny.

Especially as it's impossible to tell whether jvanderbot's "people" datagram got dropped by some intermediate hop in the UDP version or was never sent. (-:

Or, does "Alexa, add peanut butter to shopping list" fall under the UDP frame size? Joke reversed.

Your wife is possibly wondering why you send everything cleartext in one single frame and don't do challenge-response or any sort of state machine resynchronization between client and server. (-:

My wife is probably wondering why I'm using her politeness as a joke on the internet

She can welcome the coming of our future robot overlords, as between you two they'll most likely spare her. :)

That's why you put enough context in each packet that nobody gets confused if the order changes or some get missed.

And some people, like me, are... what do you call TCP but without initial handshake? Like:

----

Other person: Hi, what time was that thing?

Me: Hey, 14:00.

...

Me: Hey, 14:00!

...

Me: walks in their face Hey you, the thing you asked, it's at 14:00.

Other person: Yes yes, I heard you first time!

Me: boils internally, muttering to themselves so why the fsck didn't you say so?

----

Please don't hang on "Hello", but for $deity's sake, confirm reception of messages, especially in analog communication.

Weirdly I really enjoyed MrBeast's line on this in his leaked internal production memo:

"Since we are on the topic of communication, written communication also does not constitute communication unless they confirm they read it."

https://simonwillison.net/2024/Sep/15/how-to-succeed-in-mrbe...

This is why emoji responses on Slack are so useful. You can ask others to affirmatively confirm that they have read a message and they can make the confirmation without clogging up the channel with tons of text messages.

And no “read receipt” as it is usually implemented would not cut it, because it only means “this message showed up on the recipient’s screen once” not that they have actually read it.

I think both methods are TCP in the sense that you’re getting some form of ACK. Maybe „some people have a larger MTU“ or „some people are jumbo frames“. Hmm, but that doesn’t sound great… „to each its own MTU.“ nah.Ok, I give up =P

I just drop malformed packets that only contain “hello”

See, if you had Siri you'd be forced into the first anyway:

"Alexa add peanut butter to shopping list" "OK, what should list should I add to?" "shopping list" "Ok, what should I add for you" "Peanut butter" "Ok, playing Peanut Butter by the Royal Guardsmen on a HomePod you forgot you had"

My whole family uses Siri—both through our phones and through the HomePod mini in the kitchen—to add items to the shopping list.

Siri occasionally misunderstands the name of the item, or needs to ask who's speaking (when on the HomePod), or has trouble because the phone of the person asking has briefly dropped off the Wifi, but in the ~5 years we've had it, I can count on one hand the number of times adding has just failed with any pattern remotely like what you describe.

Let me guess, you have an American accent?

Well, mine personally shifts frequently to British—but yes; I can see that that could cause a problem for some people, especially if their accent is not a specific one that Siri's been trained on.

As a Brit, let me tell you, there is no such thing as a British accent. Our accents vary wildly by region, and sometimes those regions can be under ten miles away, let alone the huge differences between the four home nations. My own is basically what Americans imagine an English accent to be, a bit Hugh Grant but not as posh. Siri can do the basics for me, but playing some random song when any of us have asked it to do something completely different has become a running joke in our house. It's great when the kids try to set a timer or something and it suddenly starts blaring out some very sweary hip hop.

Oh, certainly—mine is vaguely East Midlands, as that's where my father comes from. But without knowing that the person one's talking to is British, just saying "British accent" is going to be much clearer.

Very true - and when I'm in the states mine is always referred to as a British accent. Or people think I'm Australian.

For me it's 80% accurate (including sometimes I'm entirely surprised it heard above all the screaming and howling) and then 20% it's just hilariously horribly wrong.

Reply on news.ycombinator.com