Yeah - this was arguably mostly my fault (sorry!).
There's quite a bit of history here, but the abbreviated version is that the dialog element was originally added as a replacement for window.alert(), and there were a libraries polyfilling dialog and being surprisingly widely used.
The mechanism which dialog was originally positioned was relatively complex, and slightly hacky (magic values for the insets).
Changing the behaviour basically meant that we had to add "overflow:auto", and some form of "max-height"/"max-width" to ensure that the content within the dialog was actually reachable.
However this was pushed back against as this had to go in a specification - and nobody implemented it ("-webit-fill-available" would have been an acceptable substitute in Blink but other browsers didn't have this working the same yet).
Hence the calc() variant. (Primarily because of "box-sizing:content-box" being the default, and pre-existing border/padding styles on dialog that we didn't want to touch).
One thing to keep in mind is that any changes that changes web behaviour is under some time pressure. If you leave something too long, sites will start relying on the previous behaviour - so it would have been arguably worse not to have done anything.
It may still be possible to change to the stretch variant, however likely some sites are relying on the extra "space" around dialogs now, and would be mad if we changed it again. This might still be a net-positive however given how much this confuses web-developers (future looking cost), vs. the pain (cost) of breaking existing sites.
I wonder, do you think "hidden" behaviors like this discourage lazier devs from using HTML standards and just use divs for everything and reimplement default functionality with 3rd party UI libs?
Seems like a good example of depending on LLMs too much. After the first bad output the next step should have been to open inspector, not to try three more LLMs.
Browser vendors put a lot of surprising, and very poorly documented, defaults into user agent stylesheets, and they can trip you up at the most basic level, and the styling is poorly exposed in most web inspector tooling.
I ran into a nasty one with Safari’s date picker—had to go down a lot of rabbit holes to figure out I needed to finesse the `::-webkit-date-and-time-value` CSS to get an input field styled correctly.
I've found that the "Computed Styles" section catches most weirdness quite easily. Remove your own styles, view Computed and you'll see "User agent style sheet" as the source.
Even if you don't look at the computed styles, the user agent styles are shown in the default element style view, at least in Firefox. There, you can see all of the applied styles as a stack, so you can see which styles have been applied first, which ones have been overridden, etc.
This is a good place to look if you're trying to figure out why an element is behaving weirdly, because it shows all the styles and exactly how they override each other, as well as the source of each style.
Something I discovered while changing the thumb track of a range slider just the other day, is if you add padding-top to the thumb, it will also add padding to the bottom.
Even better, you have to set its appearance to none for chromium before you're able to style it in any capacity, which probably makes sense to somebody but not to me.
I'd looked in the inspector after the first time I saw the gap (just added a note about that to the document) - my problem was that I didn't realize Firefox doesn't show you the default user agent styles, while Chrome does.
That's how <b> tags are made bold etc. There's a css file that's basically loaded before everything else to apply those default styles. Here's WebKit/Safari's for example:
Thanks! I had no idea. I have an issue with a large gap on mobile I should dig into. Maybe this will help me figure out how to fix it. Also just a neat fact to know regardless
> The spec lives in GitHub in a 7MB source file which is too large to view through the GitHub web interface
It's a bit offtopic, but think about it: cannot view the doc despite its size is only 7Mb ~ about 0.04 percent of a typical computer's memory (16Gb). We can open and view PDFs much bigger and search there without an issue in browser. We can view movies worth half of the memory.
But it's surprisingly not possible to view a 7Mb text file in the professional tool for working with textual source code called github. Shame. And sad.
The file is 142k lines line that GitHub would have to turn into HTML and the browser would have to render immediately.
The browser is perfectly able to render the raw file (you can try it yourself) but everything that goes with it would be asking a bit much, particularly because the intent of the author was to blame the file, therefore adding yet more information/elements to the page.
While browsers nor GitHub are beacons of performance, I don’t think it's fair in this case.
PDFs can be rendered page by page and each page is self-contained. Same with videos, where you read the index and jump to the key frame you're onto, reading data in a highly optimized and parallelized way (often down to the chip). With HTML you must render every letter that comes before line 100000 before rendering line 100000
> The file is 142k lines line that GitHub would have to turn into HTML and the browser would have to render immediately.
GitHub's new, React-based code viewer aggressively virtualizes the rendered text. Only the lines visible in the viewport, together with 10-20 lines around the viewport, will be rendered. This is why search is broken when using the native browser search (not all text is rendered).
As another comment pointed out, VS Code is capable of opening such large files with syntax highlighting, minimap, etc. VS Code is also employing some form of virtualization. It's unclear why GitHub's file viewer refuses to show files larger than a megabyte or so.
> GitHub's new, React-based code viewer aggressively virtualizes the rendered text. Only the lines visible in the viewport, together with 10-20 lines around the viewport, will be rendered.
GitHub does not load VS Code on a page load, so you can't expect it to have the same performance. Evidently they themselves know it can't handle that file efficiently.
I'm sad that the No-React forces within GitHub (presumably) eventually lost that battle. There's React all over GitHub now and it's resulting in so many weird glitches like this.
The file viewer is especially strange. Scrolling is janky on iOS (presumably they are trying to control scroll position and this interferes with Safari's scroll bar). Last time I looked into the file viewer, I believe it was rendered as a web component with its own shadow DOM and a react root inside.
Although off-topic for this thread, one minor annoyance is how the "improved" Issues interface no longer displays pagination controls until several megabytes of JavaScript finish loading. Meanwhile the Pull Request page, one of the oldest parts of the GitHub interface, is capable of displaying pagination controls before JavaScript loads, and it's interactive (i.e. clicking on pagination buttons will work as expected, even without JavaScript).
Want to note that web-based UI doesn't have to be this way. In the case of rendering text with syntax highlighting, VS Code shows that it can be done in a performant manner. As for functionality without JavaScript, embracing standard web APIs, etc., that is exactly what frameworks like React Router (formerly Remix) are designed for.
I'm not surprised to learn anything because I worked at GitHub before.
GitHub is not VS Code. GitHub is a Ruby on Rails website with some React sprinkled on. You can't expect the same performance optimization from a website that a whole IDE has.
This is also one of the use-cases I built the Tachi Code browser extension for. It injects a monaco-based editor into the current page when it detects you're viewing a raw file (i.e., the page's rendered body contains nothing other than a <pre> tag).
The fundamental difference is a dialog isn’t forced to be addressed by the user at any time, and lacks the alert dialog[0] role, at least I don’t see any of this in the example code.
I know this seems like pedantic criticism however my hope is to spread some information about the (very real) differences between a dialog and a modal, even though the dialog API doesn’t do a good enough job distinguishing this
[0]: The alertdialog role is to be used on modal alert dialogs that interrupt a user's workflow to communicate an important message and require a response
My goal here is absolutely to use the dialog element as a modal - I thought that was what it was for, and that using it in this way has significant accessibility benefits.
It is what the element is for. It can be used for both a dialog or modal UI widget. And it can significantly streamline accessibility.
MDN has a good overview of the aria attributes associated with modal[0][1] and some more explainer on the dialog API[2]
The long and short of it though is modal dialog is generally expected to halt user progression until dealt, where as a non modal dialog does not halt user progression and can be safely ignored from a user perspective.
The differences are definitely subtle. It’s also why the dialog element has .show and .showModal as an API, they aren’t synonymous, they actually set different contexts implicitly for accessibility reasons.
This is in part some of the criticism that people have of the dialog element, FWIW. There was a debate prior to its introduction to break these two UI patterns into separate elements, one being <dialog> and the other being <modal> so their semantics could be obvious and more easily preserved.
I believe the additional complexity wasn’t worth the cost but it does leave this confusing situation about modal vs dialog around.
To the create of the committee though it does go a long way in addressing both concerns, even at the cost of “obviousness”
So for the https://tools.simonwillison.net/side-panel-dialog example - where clicking on an item opens up a "details" view in a side panel, which can then be dismissed - is the way I'm using dialog appropriate? If not, how should I use it instead?
It looks to me like you’re using the dialog API well enough, but your background content isn’t inert, I can still scroll it, and if I wasn’t on mobile I’d see if tab would break the modal sandbox, as I suspect it might.
You should try using this with a screen reader if you’re able, that’ll let you know where you may need further improvement.
Right now the behavior on the page is that of a dialog and you’re using the (correct if you want a modal) modal API, that will cause issues for users
You should also use the autofocus attribute if you want the modal to focus on something other than the close icon in the corner upon opening
> I also found out today that the HTML Living Standard is very much a living standard - the whatwg/html repo has had 12,318 commits, the most recent of which was less than 24 hours ago.
I understand their mission of capturing the messy real-world usages of web technologies and distilling them into a standard, paving the cowpaths etc etc.
But I don't see how anything can be called a standard that updates daily.
In the year 2525, if man is still alive, he may find he has a web page from an unknown time period. To read it, all he'll need to do is reimplement all standards for the previous 500 years. Version numbers wouldn't make this easier at all, somehow.
And after all their efforts to get the unknown markup into an document form, they'll convert it's contents into the dominant information format of the 26th century: PDF.
To see all commits that have touched lines 120-150 in filename.txt, and see those lines. Gives a view into a subset of lines but won't help if the code in question moved out of the line range.
In Jetbrains you select a line of code and say "show history for selection" and it parses all that for you and just gives you a history of commits affecting that specific line. It's got a built in commit browser and visual diff tool.
Not sure about a GUI way of doing this, but in the CLI, I use `git log --patch [path]` all the time. It will show you the history of diffs for that one file.
the article notes some accessibility issues, those have been resolved.
on edit: misread title as take full size of the viewport, and not full height. So not necessarily applicable but keep it up as I think it is sort of an interesting Modal strategy.
There's quite a bit of history here, but the abbreviated version is that the dialog element was originally added as a replacement for window.alert(), and there were a libraries polyfilling dialog and being surprisingly widely used.
The mechanism which dialog was originally positioned was relatively complex, and slightly hacky (magic values for the insets).
Changing the behaviour basically meant that we had to add "overflow:auto", and some form of "max-height"/"max-width" to ensure that the content within the dialog was actually reachable.
The better solution to this was to add "max-height:stretch", "max-width:stretch". You can see the discussion for this here: https://github.com/whatwg/html/pull/5936#discussion_r5136422...
The problem is that no browser had (and still has) shipped the "stretch" keyword. (Blink likely will "soon" - https://groups.google.com/a/chromium.org/g/blink-dev/c/SiZ2n... )
However this was pushed back against as this had to go in a specification - and nobody implemented it ("-webit-fill-available" would have been an acceptable substitute in Blink but other browsers didn't have this working the same yet).
Hence the calc() variant. (Primarily because of "box-sizing:content-box" being the default, and pre-existing border/padding styles on dialog that we didn't want to touch).
One thing to keep in mind is that any changes that changes web behaviour is under some time pressure. If you leave something too long, sites will start relying on the previous behaviour - so it would have been arguably worse not to have done anything.
It may still be possible to change to the stretch variant, however likely some sites are relying on the extra "space" around dialogs now, and would be mad if we changed it again. This might still be a net-positive however given how much this confuses web-developers (future looking cost), vs. the pain (cost) of breaking existing sites.
Sorry!
reply