Module talk:Message box: Difference between revisions

By /

 

Line 120: Line 120:

::My issue with lightening the color is indeed that the original color is used in the context of at least one deletion mbox, and we want those to be annoying enough to be annoying.

::My issue with lightening the color is indeed that the original color is used in the context of at least one deletion mbox, and we want those to be annoying enough to be annoying.

::WCAG 2 AA should be taken a little less interestingly these days with the development of WCAG 3.0 which uses Accessible Perceptual Contrast Algorithm (APCA) to decide whether some content is too much or too little contrast. I expect blue/dark purple on pink to do better with the APCA than it does with WCAG 2 computation. [[User:Izno|Izno]] ([[User talk:Izno|talk]]) 03:13, 1 October 2025 (UTC)

::WCAG 2 AA should be taken a little less interestingly these days with the development of WCAG 3.0 which uses Accessible Perceptual Contrast Algorithm (APCA) to decide whether some content is too much or too little contrast. I expect blue/dark purple on pink to do better with the APCA than it does with WCAG 2 computation. [[User:Izno|Izno]] ([[User talk:Izno|talk]]) 03:13, 1 October 2025 (UTC)

:::Thanks for the interesting pointer to a new contrast metric! Based on a quick reading of some sites that came up in search, it seems that the WCAG working group is still considering if it will include APCA in the new release of its standards. Nonetheless, it does seem that improved ways of measuring contrast are under consideration. [[User:Isaacl|isaacl]] ([[User talk:Isaacl|talk]]) 04:55, 1 October 2025 (UTC)

For Module:Message box/configuration
Update icons to use the new Codex icons, I’ve colored them manually in c:Category:Wikimedia Codex icons with color, using the colors we used with near identical OOUI icons.

  • ambox.speedy

    ambox.speedy

  • ambox.delete

    ambox.delete

  • ambox.warning ambox.content

    ambox.warning ambox.content

  • ambox.notice

    ambox.notice

  • or this for ambox.notice

    or this for ambox.notice

  • ambox.style Not Codex, but OOUI for style icon...

    ambox.style Not Codex, but OOUI for style icon…

etc. In some cases cmbox will use different icon. waddie96 ★ (talk) 12:06, 10 August 2025 (UTC)[reply]

NOTE: The current icons are licensed such that they do not need the link to work so that arribution works. These new icons are CC4 licensed, so attribution would be needed and the link= attribute would also need removing from the code if these are to be used. — WOSlinker (talk) 12:23, 10 August 2025 (UTC)[reply]

@WOSlinker Sorry, I didn’t do the mass upload to c:Category:Wikimedia Codex icons with color, I requested it at Commons:Batch uploading/Wikimedia Codex icons, and I requested it be made with the appropriate license which is {{MIT License}} found in the gerrit:plugins/gitiles/design/codex/+/refs/heads/main/packages/codex-icons/LICENSE. Also in the package.json it says "license": "MIT".
Also some the icons have already been there for a few years, and marked with {{CC-by-4.0}}, mistakenly too I assume. Anyway, I marked my colourings of the icons with that too since it was a derivation.
Anyway, CC-by-4.0 is incorrect license. {{MIT License}} is the correct one. I will re-tag the ones at c:Category:Wikimedia Codex icons with color and all the ones at c:Category:Wikimedia Codex icons but the second one may take quite a while. waddie96 ★ (talk) 13:08, 10 August 2025 (UTC)[reply]

MOS:PDI says that a blank |link= param – which is what we essentially do with mboxes – should not be used with CC BY-SA, GFDL, or similarly licensed images. I think that I’m right in saying that c:Template:MIT License falls within “similarly licensed images”. —Redrose64 🌹 (talk) 13:24, 10 August 2025 (UTC)[reply]

@Redrose64 But the same icons are implemented in the cdxMessage component in Codex, and this component is deployed on this very wiki. This component shows the exact same images (just not using our source which is Commons) from the same Gerrit repo to display or messages we see all over. The entire package/library has the GNU General Public License see gerrit:plugins/gitiles/design/codex/+/refs/heads/main/packages/codex/LICENSE. waddie96 ★ (talk) 13:49, 10 August 2025 (UTC)[reply]

See Special:Version: codex-design-tokens GPL-2.0+, codex-icons MIT, codex GPL-2.0+, wikimedia/codex GPL-2.0-or-later waddie96 ★ (talk) 13:50, 10 August 2025 (UTC)[reply]

@Redrose64 Also, is the use of the predecessor OOUI icons with MIT license in violation of copyright at mw:Module:Message box/configuration for example. waddie96 ★ (talk) 14:29, 10 August 2025 (UTC)[reply]

Do we have consensus for these, has this change been discussed outside of this request ? Sohom (talk) 15:47, 10 August 2025 (UTC)[reply]
@Redrose64 Where in the quoted policy does WP:PDI state a blank |link= param – which is what we essentially do with mboxes – should not be used with CC BY-SA, GFDL, or similarly licensed images.? I couldn’t find it. waddie96 ★ (talk) 16:25, 10 August 2025 (UTC)[reply]

Sorry, MOS:PDI. It’s always confusing when otherwise-similar MOS: and WP: shortcuts take you to different places. —Redrose64 🌹 (talk) 16:28, 10 August 2025 (UTC)[reply]

waddie96 ★ (talk) 16:39, 10 August 2025 (UTC)[reply]

I’ve answered the edit request as  Not done for now: please establish a consensus for this alteration before using the {{Edit protected}} template. (non-admin closure) Aaron Liu (talk) 02:21, 11 August 2025 (UTC)[reply]

Could this thing use <div style="display:table;"> and <div style="display:table-cell;"> instead of literal tables and table cells (<table>, <td>)?

Were the any previous attempts to remake this using <div>? If so, what was the consensus? Sapphaline (talk) 09:46, 20 August 2025 (UTC)[reply]

A long time ago this would have been converted to divs but apparently IE sucked (this should not be news).
Today, I have User:Izno/Sandbox/Ambox with what are some skimmings and started a sandbox at Module:Message box/div. I have just done a very bad job of finishing this work. Izno (talk) 22:09, 24 August 2025 (UTC)[reply]

Which might be closer to done than not, honestly. I think what it currently needs is to

  1. Double check the ambox work
  2. Add “soft” support for the other kinds of boxes (where soft is defined as module-supports but config disabled)
  3. Merge tested /div version but not CSS into main
  4. Enable boxes of each kind one by one.
  5. Fix bugs that are identified
  6. Move CSS over to better-named stylesheet
  7. Delete the /div CSS (I won’t be too sad)
  8. Remove main support for table versions, possibly rename some things
Izno (talk) 22:42, 24 August 2025 (UTC)[reply]
I have now put this plan in a publicly editable place at Module:Message box/div/doc. IznoPublic (talk) 00:26, 26 August 2025 (UTC)[reply]
Hi! I didn’t know you were doing this, so I made something similar. It seems to work:
Iniquity (talk) 16:57, 11 September 2025 (UTC)[reply]

Yes, I did it in its own sandbox because it’s a longterm enough project not to be done in the main sandbox, which should generally be releasable ad hoc for trivial issues. Izno (talk) 17:03, 11 September 2025 (UTC)[reply]
And also, there are enough other dependencies that “just change the two or three places” isn’t going to fly. Izno (talk) 17:06, 11 September 2025 (UTC)[reply]

Where else do you think this will affect? ​​It seems that only the mobile version is oriented towards table.ambox, which can also be corrected by CSS Iniquity (talk) 17:09, 11 September 2025 (UTC)[reply]

I have already documented in Module:Message box/div/doc multiple places that will need careful work, ambox among them. Please feel free to peruse and ask questions about what’s there and why. Izno (talk) 17:41, 11 September 2025 (UTC)[reply]

Thanks! Iniquity (talk) 17:47, 11 September 2025 (UTC)[reply]
What’s broken here? I’ve probably wrong but AFAIK for accessibility the existing role=”presentation” should suffice. Aaron Liu (talk) 23:21, 28 August 2025 (UTC)[reply]

You should always endeavor to use the default aria role element. In this case, that’s not a table. Ignoring that, tables are shit for doing other presentation with. Izno (talk) 03:20, 29 August 2025 (UTC)[reply]

This /div work is great work, thanks for this waddie96 ★ (talk) 19:49, 8 September 2025 (UTC)[reply]

The elements with role=none (synonym for presentation which is not recommended for use) is not be hidden from the accessibility API but their implicit semantics are hidden. Meaning the contents of an Ambox for instance will not have the implicit semantics of a HTML data <table>. So the content given in |text= will simply be presented to the accessibility API technology as <div>Content</div> instead of <table>Content</table> which is what we want, yes. However, the accessibility technology, i.e. screen reader, etc., will not have any idea what this text is or what it represents since the <div> element that it ‘sees’ has no implicit role, such as with <p>Content</p> for instance, and there is no aria-label given… So I suggest we give it these values, especially since:

If presentation or none is applied to a <table> element, the descendant <caption>, <thead>, <tbody>, <tfoot>, <tr>, <th>, and <td> elements inherit the role and are thus not exposed to assistive technologies.But, elements inside of the <th> and <td> elements, including nested tables, are exposed to assistive technologies.

— taken from [3].

I interpret this as meaning that the <img> element inside the first <td> element will be exposed, and it is simply an image/icon with no benefit to the screen reader or the visually impaired… so give the image <td> element with class mbox-image the HTML attribute: aria-hidden="true"; so that it is completely hidden from the accessibility technology.

The next <td> nests our mbox-text class with the actual stuff we want to expose. But we want to give it meaning too. The role should be: role="note" as it’s a note semantically speaking and does not fit any other ARIA role, and the child elements of that will all be exposed to the API as whatever has placed in the text parameter anyway (as confirmed in the quote above). Finally I suggest giving it an aria-label="Notice" if |type=notice or style or content, and possibly aria-label="Warning" if delete or speedy. So it’ll read out with a screen reader for example as: Note. Notice. Content...

Option 2… the parent element may actually be what we need to expose because it’s nesting the whole dang thing? I don’t think it matters since the parent elements of the text element are hidden from the API but another solution possibility could be setting the entire <table>‘s role to role="note" and giving it an aria-label="Notice", and then still hiding the image, and then leaving the text element alone, however I’m unsure if the note role is inherited by child elements like the none/presentation role is.

waddie96 ★ (talk) 00:19, 29 August 2025 (UTC)[reply]

How about we wait to suss out how we want to deal with any accessibility questions here after I’ve got everything in divs, as in the section just above. Anyway, I think the last time I looked at this, the most similar was role=note for what message boxes represent. I am pretty sure I am not a fan of adding labels here, which should be when the content needs naming, and this does not. Izno (talk) 03:24, 29 August 2025 (UTC)[reply]
(In fact, the div version already removes the presentation role.) Izno (talk) 05:18, 29 August 2025 (UTC)[reply]

Cool. Please do ping me in discussions when this is revived. I’ve read extensively for many hours for something else and I’m quite well versed in it now. Yes, role="note" is the most appropriate role. Also, after testing it in the enwiki context using a few different screenreaders I tend to agree with you that an aria-label in this case is unnecessary “fluff” to the accessibility tree for the mbox, since as you stated the text describes itself. 👍 waddie96 ★ (talk) 14:56, 29 August 2025 (UTC)[reply]
Awesome. Can it give the image the HTML attr aria-hidden="true" too? Please 🙏 waddie96 ★ (talk) 14:59, 29 August 2025 (UTC)[reply]

The discussion above is closed. Please do not modify it. Subsequent comments should be made on the appropriate discussion page. No further edits should be made to this discussion.

Wow, thought I’d summarise:

  1. Change the ARIA role given to the parent <table> element to role=none. Synonymous to presentation, i.e. identical function of removing all semantic meaning of table element and its direct children, but the latter is soon-to-be deprecated HTML.
  2. Hide the images please: wrap the default image, and especially the custom left and right images (since the default at least has a null |alt=) with <span aria-hidden="true">...</span>.
Sorry, I’d edit the module in sandbox and give diffs but I have no idea how to code in Lua. Thank you! waddie96 ★ (talk) 19:14, 2 September 2025 (UTC)[reply]

I think Izno suggested waiting until the div conversion was done, and then looking to see if anything like this is needed? — Martin (MSGJ · talk) 14:33, 4 September 2025 (UTC)[reply]

Yes apologies. I see that now. waddie96 ★ (talk) 23:44, 4 September 2025 (UTC)[reply]

Current background color doesn’t meet WCAG AA contrast requirements with links color used in Vector-2022, Timeless and Minerva, so I request to change it from #ffdbdb to #ffe6de.

More specifically, edit the 14th line so that 

	background-color: #ffdbdb;  /* Pink */

becomes 

	background-color: #ffe6de;  /* Pink */

. Sapphaline (talk) 18:02, 10 September 2025 (UTC)[reply]

Hmm why not change it to var(--background-color-destructive-subtle, #ffe6de);? waddie96 ★ (talk) 18:05, 11 September 2025 (UTC)[reply]

Then it’ll be dynamic, adaptive for CSS themes, and change in dark mode waddie96 ★ (talk) 18:07, 11 September 2025 (UTC)[reply]
Why did the WMF feel the need to mess with link colors again? (Sent from the REAL vector skin) —pythoncoder (talk | contribs) 17:57, 16 September 2025 (UTC)[reply]

Hahaa I +1 this. waddie96 ★ (talk) 18:30, 16 September 2025 (UTC)[reply]
Not done for now: This is also the background used for cmbox speedy and delete. I’d be fine making this change for all or none. I also think it’s a reasonable question whether to consider the token directly for this case. Izno (talk) 23:37, 30 September 2025 (UTC)[reply]

As for the token, the actual token is #ffe9e5 in use with background-color-error-subtle, which is probably the preferable token to reference rather than destructive-subtle.
My issue with lightening the color is indeed that the original color is used in the context of at least one deletion mbox, and we want those to be annoying enough to be annoying.
WCAG 2 AA should be taken a little less interestingly these days with the development of WCAG 3.0 which uses Accessible Perceptual Contrast Algorithm (APCA) to decide whether some content is too much or too little contrast. I expect blue/dark purple on pink to do better with the APCA than it does with WCAG 2 computation. Izno (talk) 03:13, 1 October 2025 (UTC)[reply]

Thanks for the interesting pointer to a new contrast metric! Based on a quick reading of some sites that came up in search, it seems that the WCAG working group is still considering if it will include APCA in the new release of its standards. Nonetheless, it does seem that improved ways of measuring contrast are under consideration. isaacl (talk) 04:55, 1 October 2025 (UTC)[reply]

Must Read

Leave a Comment

Your email address will not be published. Required fields are marked *

Scroll to Top