21 comments

[ 0.24 ms ] story [ 49.0 ms ] thread
(comment deleted)
You can click the "eyeball" icon in the top-right to expand every post at once.
fuck i got all the way to the bottom expanding them one by one and then tabbed over and read this. is it some practical joke? why not just label it
(comment deleted)
Labels are bad.

Just a barely visible icon off to the side with no description is the way.

A certain irony on a accessibility discussion thread.
(comment deleted)
Any reason why everything is "hidden" by default? It's imho a very unfriendly UX (which is quite ironic coming from a thread criticising "GUI")
This decision is made by the author of the post in question.
I'm surprised GTK/GNOME has had issues with accessibility deteriorating; it always seemed like they had the best story for it, and I expect Fedora and Ubuntu to care about it. Unfortunately, it's hard to be surprised that if anything would deteriorate, it's accessibility, either.

But while I understand it's not the point of this thread, which is probably mostly venting, maybe it'd be a good idea to focus here on trying to find what can be improved. I'd definitely be willing to sponsor work on improving accessibility in Wayland and Linux in general, patron-style. It'd be nice to have serious effort put into the protocols the way that we are seeing attention put into color profiles and HDR right now.

I believe we need to bypass all of these bespoke accessibility solutions and use machine learning to visually parse a live feed of your computer screen. We should be able to take a given screenshot and determine where the buttons are, what the text is, and make something that is decently workable. Then we won't need to care about each app, each website, each operating system supporting accessibility standards because our accessibility tools will be using the computer in the same manner that humans do, visually decoding a live feed of images.

Easier said than done, but given what is happening in AI right now, it doesn't seem far fetched to think that we'll be able to accomplish this.

I was hoping to do something like this for adventure game hotspots after a hand injury. Even experimented with AutoIt and its OCR, before giving up.

Now there is another niche project with the goal of translation: translate.net

With more capable models it's going to be fun to see what folks come up with!

Still, I hope internal efforts get some attention as they don't need to rely on hacks and awkward workarounds.

It doesn’t sound unachievably hard, but I cry for all the “wasted” energy it would require.
From a user perspective (i.e., not a developer one), what have either of these projects improved since the old days of GTK2 and X11? Maybe I'm not enough of a power user, but I see no tricks in my Linux GUI today that weren't already there 15 years ago. Preemptively, I reject the "security" answer, since Intel/ AMD/NSA/Mossad can already peek my RAM at will.

The whole mess feels like corporate sabotage.

In the beginning of Gnome 3 I loved this feature where you could reply to someone’s chat messages straight from the notification without switching apps/windows: <https://lh6.googleusercontent.com/-hIyeIWfFyvc/ToRSxqZnFQI/A...>. It eliminated a costly mental context switch.

But XMPP is dead, so I don’t use such apps anymore. I also haven’t been using Gnome for a very long time now. I’ve read somewhere that they removed this feature though.

(comment deleted)
XMPP isn't dead, but quite active: https://xmpp.org/blog/

It's very possible your social circle is not on XMPP though. There's some light at the end of the tunnel with the EU's Digital Markets Act forcing the large messaging providers to provide interoperability - at a minimum this should enable better bridges between XMPP-based and proprietary systems.

The active GTK clients these days are Dino and Gajim. Pidgin is still around, but has fallen very far behind with regards to features (they're in the middle of a rewrite that will hopefully address this).

Peeking at RAM and literally being able to record your whole screen from an npm package are two quite different levels of threats.

Also, multi-monitor setups with different DPIs are not possible on X.

Someone who has a malicious npm package running on his machine has bigger problems. It will already be able to read/write every file that the user can. Why is the screen so special that we need to break everything?

I see so many screencast/input-sharing/macro apps on github that have been in a holding pattern--for years!--waiting on wayland to either fix a bug or add a feature. 14 years of wayland so far, and we still can't do things that every other major OS's display server handles easily. At this rate, wayland will certainly become exploit-free, since no one will be using it.

Gnome/GTK accessibility has been deteriorating for years. Unrelated to Wayland.

Hamburger menus that cannot be navigated with Alt+_ shortcuts (and sometimes not even with cursor keys) are the prime example.