Comments by "Mikko Rantalainen" (@MikkoRantalainen) on "Theo - t3․gg" channel.

  1. The real magic: accept 2h delay for every change and you can cache everything for 2h.
    70
  2. 12:15 The idea of <picture> element is to allow creative decisions. Some big publishers like to swap images in the content depending if you're using widescreen display or smartphone in portrait mode and the <picture> element can do this without any JS. It also supports defining separate images for HDPI-displays vs normal monitors without using JS. If you're not interested in any of that, <picture> has very little value for you.
    66
  3. This is about computing it all client-side so that you don't have to pay for server resources. And you assume user-base is too stupid to run CLI programs so you would need to run imagemagic on server resources.
    5
  4. The <strong> and <em> are more recent additions than <b> and <i>. I think the idea was to make content more accessible for screen readers or such but in reality those are just longer spellings for <b> and <i>.
    3
  5. 23:00 The cause of this difference is "due historical reasons" as with everything in life. In fact, the "disabled" alone is not a flag but it's an SGML (of all things!) shorthand notation for value "disabled" if exactly one attribute allows value "disabled" (yes, SGML is truly this brain-damaged). Technically it's shorthand for writing disabled="disabled" which is also why you have to use the latter syntax if you write XHTML instead (XHTML being implemented in XML, it obviously doesn't support SGML shorthands).
    3
  6. 15:20 I don't understand this argument. If you're happy to store 32 bit integer in binary form, you should be happy to store the 128 bit UUID in binary form, too. The database can convert from binary value to human readable value on the fly, just the way it happens for binary 32 bit integers, too! With the same logic you could argue that 128 bit binary UUID is actually better format than 32 bit integer stored as base-10 number as a string. That said, if your database doesn't support UUID column type, maybe you should switch database.
    3
  7. 24:00 I agree 100%! I think "off" shouldn't even be a keyword for the exact reason you explain here.
    3
  8. Or some people simply like the blur effect. Just like some people think that having the room lighting to sync with the contents of screen content on your TV is a nice thing to have.
    3
  9. If you only know click-and-point, CLI seems too complex. And I typically use "convert" instead of "magick".
    2
  10. 19:40 You still have to look at the generated assembly if you want to get the runtime from 40 seconds to less than 1 second!
    2
  11. Otherwise great video but I would say I hate <marquee> with passion as a reader of any web page that uses it. As a result, it cannot be S tier.
    2
  12. I don't consider this as a failure of open source. It just another example of XKCD 2347. Everybody using a piece of software some lone person maintains as a hobby, maybe. This happens with closed source software, too, but the problem is hidden beneath the surface there.
    2
  13. 33:25 Always look at the total wall clock time. User time ignores all memory allocations and IO which are important for this kinds of tasks.
    2
  14. I think V8 would be more accurate to explain as V4 + 2 extra random bits. V4 has 122 random bits and V8 has logically 124 random bits. V7 is V1 with opaque/random node identifier + no offset to UNIX epoch.
    2
  15.  @dh00mk3tu  Most of the OS visible parts are just OS vendor provided apps that are launched automatically when your session starts. On Linux this is especially clear: I'm currently running "picom" as my window compositor, "xfwm4" as my window manager, "xfce4-panel" for "start menu" counterpart of Windows (includes the whole bottom panel with icons). My desktop environment is even running separate process to render the task bar buttons within the panel etc. All those processes are just regular user mode programs and if I kill one, the system may or may not automatically launch another one depending on how critical it's for the user interface. For example, if I kill the process that takes care of rendering weather in the task bar, it will not be automatically restarted until I logout and login. Or I relaunch it manually.
    1
  16. 33:50 CSV with semicolons is just Microsoft variant of CSV. The use C as in semiColon.
    1
  17. 21:05 I think the reviewing code is the most important part. Most of the code that's actually maintained in long term will be read many times more than it's written (computer effort doesn't matter for this discussion, only human effort). In addition, the code will be read by other people but the original author so it's about communication. Because using const can help readability and understandability of the code, you should use it for those purposes. However, don't expect to see similar safety benefits that C++ can provide with const, where const reference to object is honored recursively in function context. For write-once toy projects where the code is never used again for anything important, sure, use "let" for everything. But don't forget that nothing is as permanent as a temporary solution that seems to work for now.
    1
  18. 12:15 I would argue that anything that is performance limited at 500 iterations/sec is not that good a result. It might be better than competition but considering the amount of work a computer can do in 1/500th of a second, that one iteration must be insanely expensive to run.
    1
  19. 4:00 You could say "If Safari sucked less" about a lots of things...
    1
  20.  @virtual5754  Modern browsers also support Oklab which allows pretty painless color definitions.
    1
  21. 7:30 You can split the whole repository into submodules (each with their own git repo) and just use "git fetch --recurse-submodules" or set the respective config. However, Git submodules are also a bit of a pain to work with. Our team uses those for some things but it's not a feature you should be mindlessly using. In the end, I think you should think hard about what's the project you're working in. I think even Linux kernel is too huge project for any single human to fully understand. That kind of thing shouldn't be a single monolithic project anymore because nobody cannot fully figure it out anymore. Instead, you should create understandable interfaces between components and keep each component small enough to be understandable by humans. And yes, once you create interfaces between components, if you make a mistake while defining that interface, fixing the interface will be much messier if you use submodules. This is the same problem as using microservices in general, except that microservices also introduce additional latency, no matter what you do.
    1
  22. 5:10 From the information I've seen I've understood that there was not logic in the update files, just data structures. But Crowdstrike kernel mode driver had such a low implementation quality that it doesn't verify the data that is passed in from the user mode or filesystem. As far as I know, this is an absolute no-no in kernel mode programming so the true cause of this failure is shoddy implementation by Crowdstrike kernel developer team. It's also a nice demo how little benefit WHQL certification can actually provide.
    1
  23. You looked like a doubleganger of Jimmy Broadbent in the thumbnail of this video.
    1
  24. This would have been about 10x simpler if you hadn't used React.
    1
  25. It's still too early to tell for sure but it appears that the update file doesn't contain executable code and the actual bug is totally missing input validation in the kernel driver. I would expect that they do sign the update files and blindly trust the contents of the update file. (As long as the signature is correctly signed and correctly verified, verifying the update file on server side should be equally safe to doing the file verification in the kernel driver.) (1) However, assuming the update files were actually signed, the server didn't do the verification either! (2) And if the update files were not digitally signed, this poorly written kernel driver simply did zero verification for its input files! In the end, it doesn't really matter if either (1) or (2) is the reality. Crowdstrike obviously has MAJOR skill issue and shouldn't be writing kernel drivers at all.
    1
  26. A correctly implemented UUID v1 will definitely be unique for all machines correctly following the spec. However, correctly implementing UUID v1 requires real synchronization within one system and many programs that want to create UUID v1 identifiers are not synchronized to each other. In addition, it requires using real non-duplicated MAC address. In fact, in many implementations v1 identifiers are not synchronized even between threads in a single process because actually correctly implementing the full UUID v1 spec is hard.
    1
  27. 1:15:00 I think this is a great illustration of the current situation and the web components definitely belong to the left side of the spectrum just as you put it. However, we need some new standards to live around ShadCN level or even slight closer to "component I wrte in my codebase" level but we don't have anything yet. And it's still unclear to me if there can be any non-framework agnostic solution here. Whatever solution is chosen, I hope it can fully interact with DOM. If you have any kind of virtual DOM solution where a piece of JS code overwrites the actual standard defined DOM from virtual DOM, that's definitely a model that cannot co-exist with standards compliant code. We would need a solution where you can have some server side rendered components co-existing with code that uses native JS code to work with the DOM without both sides stepping on the toes of the other side.
    1
  28. For a big systems, the rationale for keeping closely generated ids close to each other in the index. In reality, the generated ids are used in context with the surrounding content or data. The ids generated nearly the same time in a big system are only rarely related to each other so it makes zero sense to keep those close to each other in the index. (Imagine something like Gmail. Do you think it makes any sense to keep all the mail received near 2024-05-01 13:00:00? How often do you need to go through all the mail received near that time in the future?)
    1
  29. It seems to me that web components (WC) are great if you're trying to re-implement something like `<input type=datetime>` but it's a poor fit for server side rendering (SSR). Unfortunately, many JS authors try to use web components for SSR. Browsers really would need to support SSR better but the current state of WC is not designed for that.
    1
  30. 7:20 Note that PageSpeed Insights shows emulated mobile performance results by default. The performance metrics are generated from emulated "Moto G Power" with throttled slow 4G connection. Switch to "Desktop" tab to see more representative numbers when you're not limited by the client device. Unless you provide more limited content for low-power mobile devices, PageSpeed Insight will show lesser results for "mobile" clients.
    1
  31. "There's Only XUL"
    1
  32. I think masonry should be optional feature of either display:grid or display:flex. I would personally prefer flexbox because masonry layout seems more like flex layout than grid to me. If two fully independent browser engines (e.g. Gecko and Webkit, instead of Webkit and Blink) have already usable implementation for display:grid based solution, I'd go with it already because I want this ASAP. That said, I understand the argument that Google is making here and it does seem the best way forward in long run if vendors accept having to write more code for the initial version. And the ability to use `auto` sized tracks with the Google proposal seems like a very nice feature to have.
    1
  33. 1:17:05 I think the fact that web components do not behave similar to native components (e.g. cloning) is the issue here and that part should be fixed in the web standards. It doesn't matter if the difference is hurting framework's internal implementation or native JS code you're trying to write by yourself. If this is caused by specs, then the specs would need to be improved to fix this and make web components to work when a fragment of DOM is cloned, or at least immediately throw instead of silently failing to clone parts of the fragment. Changing some old APIs to throw in case web components are passed in is an acceptable solution if the API has design that truly cannot support web components. Hopefully the error message would prod towards the newer API for the same use case that does support web components. Every API that has to be broken must have newer counterpart that has minimal changes to make it compatible with web components.
    1