This is definitely another counter reset

"If your thing is small, it's already performing" I needed that.

Babe wake up, another JavaScript "bundler" has just shipped.

It's funny to see that Theo didn't know the "disable cache" and throttling options were only active while the dev tools were open. I thought it was obvious, but it's always interesting seeing ppl with obviously more experience than me ignore things I thought was obvious. Which shows how much things we take for granted/obvious can be a big source of mystery for others. Even among devs.

wait. This is how JDSL works.

If anyone from Twitch wants help with the 3pm slouch, HMU, I'm available for consulting 🤣 (And yes, the \_1 vs 1\_ thing was a typo.) Thanks for the video and your comments!

Across the TanStack projects we've pretty much moved all the packages within them to have the bundling portion of it handled by Vite (rollup). It's actually pretty neat, with a shared TanStack/config package that has the shared configuration. Router has been using the shared package since pretty much v1.

In our local dialect, Kuto means "Lice" LOL

I guess the cross site tracking could be improved by having the browser cache the most common libs? Super interesting side channel attack though.

Sounds like the goal is having the client itself act as a sort of git repository - actually sounds like a really cool idea! I have a feeling that at some point, such as with major releases, it would make sense to start anew and ship an entirely new “baseline” for which to build of of

Loved the "am I getting corrected live?" Hahhaha😂

Interesting. Very interesting. Interesting.

I remember i did a hobby project where i would pretty much store most code and assets in local storage and very minimal js would figure out how to extract and reuse or refresh anything that is missing. Ages ago when local storage was an innovation... and i'm not even a frontend dev i was just learning shit. It worked pretty well, everything felt like its static html shipped from some file server.

bro didn't say anything other than "very interesting" 😂

Very interesting

yay I was at the talk in Sydney was a fun night!

Damn. Cant believe we havent thought about doing this before.

I feel like we're zig-zagging towards (1) a service worker that manages chunks as strings and can load/discard/execute them, plus (2) an API that can deliver on-demand groups of chunks to the service worker. (Surely dozens of apps have implemented flavors of this over the years) We don't really want browsers to have to keep re-compiling functions that were only used 3 deployments ago, nor do we want these old bundles still in use causing more vital resources to be bumped out of the cache (seems a lessor concern).

I work on a giant complicated codebase with way too much repetition that is impossible to advocate for time to simply and improve core performance. This feels like it may actually be helpful to me! I wonder if it can cache across apps on the same site?

Weird. But I get the point. Used to have this issue before where the bundle would be big enough to kill first load by quite a bit, but an alternative is Qwik which we moved to and basically does the Kuto thing by default.

Your bytes-to-download are going down, but your memory & JS execution impact on the browser is going up (presumably it has to still recompile the unused functions on every page load). This could be a bad trade-off depending on whether local RAM/compute or network is the bigger bottleneck.

Very interesting x1000 :)

This kind of reminds me of Snowpack. HTTP2 (and now 3) allows for better performance for many files, so just make every dependency its own bundle. They rarely change, so the only cache eviction is your application code. Although this is more granular I guess

I'm confused. Why are we still bundling code? Shouldn't stuff such as http2 solve the issue of having to download a whole bunch of tiny files?

13:40 Caching pretty much makes everything better.

At first when it said you have to ship everything again, that didn't make sense until right at the end when it pointed out it's when using a single bundle. I use multiple bundles split on route's & dynamic imports, and doing an update is usually very small, unless say I updated a big lib like React etc. So for large projects with lot's of routes & lots of dynamic imports this is maybe not for you. But were this might be useful is for Web Games, where pretty much everything needs loading at startup.

15:45 could have used this clip as teaser at the beginning of the video. Succintly summarizes what Kuto does

The reference to Schrodinger's cat is brilliant! (Schrodinger's function)

FYI, Filipino: kuto [kú. tu.] : lice (n.); louse (n.);

I'm a chess player, and so initially I read "blunder" instead of bundler haha. It pretty much means a big mistake, so i was really intrigued how a mistake makes your site faster

Thanks for reminding me to turn caching back on haha

This sounds interesting for my work but I notice that if I drop it in as a replacement for our current process, we'll have a markedly different set of artefacts in dev than production - because production doesn't build as often as dev, it'll have different cache corpuses. That shouldn't matter in theory, but it could be super hard to debug if anything did go wrong

Dynamic import and bundler's checksum-based naming fixes most of these issues, arent they?

Kuto reminds me of a different way of module federation builds

prepare yourself for cache invalidation horror stories

God webdev still seem so archaic.. i see devs from the future laughing at us.

Seems interesting, not sure I have a project large enough for this to make sense though. At least not a personal project.

Life is good after switching to GoTH stack😌

Oofff, I read that as "A * blunder * that makes your site faster" And I went "huh not often that a mistake gives a performance boost, ... everything JS is a blund .... OOH a BUNDLER" .. Haha. My bad folks.

first load time is really important for web*sites* but you wouldn't use a complex bundler, framework and such for them, *right*? when it comes to web*apps* that you visit frequently and run extended sessions on the first load may take some time, just make it run fast (and preferably don't duplicate too much code between tabs *please*)

I think for this to be viable when using with sentry io, it MUST have a source map for the final output.

kuto means lice in filipino lol

As a non-JS developer, this seems cool, but fundamentally backwards. Imagine if JS files were shipped as separate http endpoints, then it'd be trivial for the browser to cache these properly between updates. Of course extra http requests are bad so there'd have to be a strategy to mitigate that, but that sounds much more sane than what's done here, no?

@t3dotgg when will you make a Hookdeck video?

Take a shot every time he says "interesting"

another one?

Wouldn't it be nice if browsers would handle caching with diffs like Git?

I think I have a better solution for incremental updates utilizing import maps 🤫

whats so bad about downloading a few MB? Doesn't seem worth the tech debt but maybe I'm missing something.

wow

F CACHING, yes ive said it

My first impression is, that it may improve user experience on apps that have grown and have a mostly inelastic user base but building a new app will not benefit from this.

kuto is lice in my language xd

cod in shambles after that call out

not again!

reset the timer

Unless your code is humongous - like, way bigger than a mb - I don't see the selling point.

Cool concept but way too much added complexity for little benefit within the context I work in

"Forever" in a browser is a JOKE

not again

You e plained how great it is and how it could be useful. But you forgot to criticize it and talk about its possible downside. You also forgot to talk about any existing solution in the wild.

Cache shouldn't be a crutch for your site. This is a big reason there's so much bloat. Developers think a website should be downloading a billion different libraries with no regard for their users. Blaming the user for not wanting to fill their drives with your garbage code is just mental. Cache is useful, but blaming performance of first load on users is ridiculous.

welcome to the new era of bloating the user hard driver where every website demands 10gb of space

That jab on privacy is uncalled for, and not true at all. I get that you're comfortable handwaving it because you're a public figure of sorts, but not all of us are Nevertheless, thanks for the video

just realized ive been running without caching....

first

I like your every day videos and consistency. But here's a suggestion, please try to keep your videos not longer than 15 minutes. I don't know if its the case with everyone or just me, but my brain intercepts videos greater than 15 minutes as "long videos" and I tend to add them to watch later sometimes instead of watching them right away

please stop..just stop..

you can solve this at the code organization level. The dev should be in control of code splitting, not magic solutions.