P99 0 ms* autocomplete for 240M domain names

ddbalatero about 4 hours ago 32 commentsRead Article on ruurtjan.com

HI version is available. Content is displayed in original English for accuracy.

⚡ Community Insights

Discussion Sentiment

71% Positive

Analyzed from 820 words in the discussion.

Discussion (32 Comments)Read Original on HackerNews

skybrian•about 2 hours ago

This autocomplete suggests domains that don't exist. You can just type garbage and it will suggest something, but then if you go there, there are no records.

It seems like one purpose of an autocomplete box is help you avoid typos, so that makes it less useful.

pierrefermat1•about 1 hour ago

Yes, OP seems to have completely lost touch with what is actually useful vs optimizing metrics for the sake of it.

lnenad•6 minutes ago

But there is a cool blog post about it though.

chrismorgan•about 2 hours ago

Using keyup makes no sense and is inconsistent with user expectations. For triggering actions (which includes normal typing), you only ever use keydown. (Well, there’s one exception for reasons unclear to me: activating a button by pressing Space. That triggers on keyup like how clicks are on release, while Enter triggers on keydown.) Keyup is limited to things where you’re constantly reacting to the state of a key, as is common in games.

This affects the functionality, too. It is in fact introducing latency by using keyup instead of keydown. Feels bad.

hnlmorg•4 minutes ago

Key down is used for events that can be duplicated by holding the key. Eg in a text editor, when you want multiples of the same letters, you’d press and hold the key.

Key up is used for when it’s important to only have one occurrence of that event.

Technically you could write code that made key down only react once. But the logical separation makes some sense.

zxexz•about 2 hours ago

As a user, I’ve not thought too much about this before now. I agree with you mostly, but the keyup on space behaviour actually feels so innate I’d hate any change to it. Keydown on space is “jump”; nothing else.

chrismorgan•about 2 hours ago

I’d just like to hear an explanation of why it is. Because if you handled “is key currently pressed” events with polling rather than events (which is how people almost always consume them logically), it would be the only thing in the entire world that I can think of that would/should use keyup.

Edit: I think people are misunderstanding me. I’m asking for an explanation of why, when you have a button focused, pressing Space activates on key up, rather than key down like literally everything else on the keyboard, including Enter on a focused button. (Mouse activates on release.)

Rygian•40 minutes ago

Because in those contexts, the Space key is being used as a physical mirror of the on-screen button, and on-screen buttons are UX'd to perform action when released.

These two interactions should behave identically:

- Hover over a button, press mouse button 1, release button 1.

- Tab over to a button, press spacebar, release spacebar.

Why on-screen buttons are UX'd to behave this way is a logical follow-up question. I'd wager that it gives a means to bail out of the clicking (e.g. by moving the mouse out of the on-screen button before releasing mousebutton 1, or by pressing Esc while having spacebar still depressed).

podocarp•about 1 hour ago

It just matches real life. What about lmb to shoot? You don't fire a gun on trigger release, you fire it on trigger down. Walking? Would be weird moving only on keyup.

dangond•about 1 hour ago

Jump is keydown because many games let you adjust your jump height based on how long you hold down space after the jump starts.

oersted•about 2 hours ago

Why not just trigger the fetch on keyDown and show it as soon as the response arrives, as usual?

The time it takes to press a key is a reasonable target to aim at for API latency I suppose, but it is still an arbitrary target. Waiting to display until keyUp just adds more latency if your API is faster. Having it synced with keyUp doesn't make it feel more immediate to me.

s4i•about 2 hours ago

It’s usually good to wait for user to do KeyUp before treating the string as final. A fellow commenter explained some of the reasons: https://news.ycombinator.com/item?id=49505993

ViscountPenguin•about 3 hours ago

Unfortunately this approach doesn't feel that great down here in Australia, definitely a function of latency.

I think you could get a lot closer by framing this as an optimization problem, where you use the full alphabet dictionary, but add a residual prediction which aims to cover as much of the remaining domain name tree as possible weighted by popularity. This tree could then be pre-baked and stored with the same system. This would probably get you p99 0ms even in Australia.

kevmo314•about 3 hours ago

If you’d like to reduce the network latency further you can store each trie node as a file, naming it conveniently the prefix path to that node. Then dump the few hundred million files onto R2.

Now the traversal can be done completely via CDN lookups!

camel_gopher•about 3 hours ago

Clever but that’s not how we measure latency.

dbalatero•about 3 hours ago

When it comes to UX, perceived latency is king.

wky•about 3 hours ago

The perceived latency starts from keydown, not keyup. Redefining latency to start at keyup reduces measured latency, not perceived latency, and delaying the visual display to keyup makes perceived latency strictly worse, not better. Even sticking with the keyup definition, just displaying the result as soon as it is available gives the possibility of negative (defined) latency.

weird-eye-issue•about 3 hours ago

I think you don't fully understand. They aren't just doing a search based on the key being pressed down. They already did a search based on the previous characters that returned results for all possible next characters. So by the time you type a second character it just checks results locally from the search that had likely already been returned from when you had typed the previous character.

vasco•35 minutes ago

What you said might be true and yet what OPs AI agent did wasn't measure latency of the service nor calculated a p99 for it.

bagels•about 3 hours ago

Looks more like 500ms?

cortesoft•about 3 hours ago

KeyDown events don’t work great for mobile, though.

pupppet•about 3 hours ago

Autocomplete aside, this is a pretty nifty tool.

pixelpoet•about 3 hours ago

Pretty sure we mean < 1ms rather than actually instantaneous.

weird-eye-issue•about 3 hours ago

On a technical level yes but once you factor in monitor refresh rates then you can get to levels of optimization where it simply doesn't matter because you are constrained by waiting for the refresh rate anyways.

lovich•about 3 hours ago

Its like tic tacs saying they are 0 calories because they got the per serving size down low enough to round to 0.

ChannelFence•about 3 hours ago

its pretty clever but what happns when someone pastes a domain or uses IME or voice input? the api being that fast is still impressive.