Modern design philosophy is that buttons with labels are messy and complicated so they should be replaced everywhere by hamburgers and hieroglyphs.

I threw up a bit in the back of my mouth typing that, but it's true -- what started as spatial triage on mobile is now the universal design language everywhere, and it sucks.

Regardless, there's still aria-label.

I think they’re referring to aria-labels, not visible text.

Product owners don’t care about accessibility. They care about time to market for 99.999% of your users, or for many products 100% of the users. The products I build are not accessible. Designers work hard to ensure they’re not or quite hard to make accessible at best. Modals on modals on modals with important content changing all over the place!

Is it on a "product owner" to care about accessibility? Or engineers to build things properly? If this was any other engineering field - civil, mechanical, electrical, there would be standards which we seem to lack / fail to enforce

If the product owner directs work at a ticket level and never assigns it to anyone, it's their fault it isn't there. If the engineers have more autonomy than I've seen at a company are were able to decide what to work on, then sure blame the engineers.

> there would be standards which we seem to lack / fail to enforce

This is by design. Software developers are hardly "engineers". They build what the business wants, quickly, and worry about the "properly" after product-market fit is found (or so they're told). This dynamic would be upended if software engineering were licensed.

I can’t even convince design not to make the text grey on grey. It’s not just “disabled” users who are suffering this nonsense.

I get paid the same not to fight. More even, because I’m a team player.

The way this gets fixed in an org isn't by engineers winning fights with designers.

It's by someone at a higher level deciding that they care about accessibility (or that they care about not getting sued), and introducing constraints which their designers and engineers must work within, and making compliance with the constraints part of the QA process.

The way it gets fixed across organizations is government regulation

Relatively few engineers in the US are licensed. It's fairly common with civil engineering because at least moderately senior engineers are signing off on drawings for regulators. But mechanical/chemical/etc. engineers are not in general licensed overall--maybe 10-20% are the numbers I've seen.

And that's not even counting all the jobs like sales engineer, mud engineer (oil business), etc. that are really more engineering-adjacent jobs.

When I took the engineer in training exam--never had a reason to follow through on a PE, it basically required things like a four-year engineering degree, working under someone for some number of years, passing the PE exam, and presumably signing code of ethics, etc. For certain jobs, it is pretty much a requirement past some level but mostly doesn't matter a lot.

> worry about the "properly" after product-market fit is found (or so they're told).

Thankfully for Reddit they found the product market fit 10 years ago.

They're just a bad software company.

This is why I found it hilarious when people say they prefer proprietary software because a commercial company will produce better code than volunteer individuals.

Train drivers are also called engineers, and much software development can be seen as engineering in this general sense.

It's on product, because it's an ask from the customer (and requires planning into the roadmap - it doesn't come for free).

Wasnt there some push in the US to make websites acessible or face some very hight fines ? I do remember something like that.

It matters a lot when you have a Gov. contract of some sort.

Reddit doesn't sell to the USG or any US schools or etc.

Interesting, so that is the difference. Thanks! :)

Fortunately the ADA[1] means that you can easily reach their real customer support AKA legal department.

[1] https://www.boia.org/blog/the-robles-v.-dominos-settlement-a...

Having a site good for accessibility means it is also easily scraped, if it is easily scraped no need to pay for API access, and there goes the money grab

It is not they cant do it right, they do want to do it right because they need to monetize the site to appease the VC's,

Anything that can prevent that, including accessibility, goes

A custom program can find the buttons by label. It's just not labeled in a way the screenreader is looking for.

it is really not that much harder to scrape using XPath and no labels if you know what you're doing…but only if you can see things, of course.

Because (to oversimplify things by a lot) you don't have to care in most cases. You have legal requirements as a public institution, when selling to governments / educational institutions or as an employer (depending on country), but requiring all privately-run, commercial websites to be accessible is usually not a thing. This is not legal advice by any means, and the actual situation is way more complicated than that.

Totally agree. Whilst I think that Reddit _should_ support their blind users in this case, as a general/blanket rule that you can sue anyone who doesn't then everyone's blogs/dinky little sites need perfect/proper accessibility done for them, which just isn't feasible.

Reddit is large enough that if they don't have a positive response for that sub then they're just a total disgrace. Who TF wants to make it harder for blind people to live their lives? Only evil corpo fuckers, that's who.

sounds like they will be sued now! ADA trolling is a thing and it does catch big fish.

if you look at the regular reddit "new" interface you'll totally understand how this occurs.

the interface is a mess, accessibility concerns or not. It's truly an equal opportunity offender in all the worst case.

As a non-web-dev who sometimes builds my own toys, when you say someone building buttons without labels is lawsuit-worthy, it makes me think you're all saying something other than what I'm interpreting. I'm guessing you mean more than

    <button type="button">My Label</button>

or something like that?

That would be the label of a button yes. I think in practice, what's missing is describing the label of a button (or another element with that role) with an aria-label attribute.

Buttons that miss a label are often an icon without text and label (e.g. a delete button with a trash icon).

https://developer.mozilla.org/en-US/docs/Web/Accessibility/A...

Thanks for the explanation!

The lawsuit worthy aspect is that labels for buttons are super important for understanding what a button does. A lack of labels (either through not putting text inside the <button> element or using a technique like aria-label) can quickly make an interface unusable for a blind person - the visual equivalent would be completely blank, identical buttons. That barrier is what someone could sue over.

It can also quickly make an interface unusable for a non-blind person, if they block external fonts.

My understanding is that a button used like this is already accessible.

However a button that looks like this:

  <button type="button">
    <img src="/delete.png" alt="Delete post" />
  </button>

Is not. Screen readers will read the simple text on the button, but any descendants of the button will be considered purely presentational (even if as in the example above the img tag includes an 'alt' attribute, it won't be read).

Modern screen readers will now often parse text content out of the button even if there are descendants, but this behaviour is not reliable or according to spec.

To fix:

  <button type="button" aria-label="Delete post">
    <img src="/delete.png" alt="Delete post" />
  </button>

There are other techniques as well such as using CSS to hide text (not with 'display: none' though, which also hides from screen readers, meaning this often feels a bit hacky) or using a hidden element with text content and aria-labelledby to reference the element's ID.

This is a pretty good walk through: https://www.sarasoueidan.com/blog/accessible-icon-buttons/#a...

No, your first code snippet is totally fine. The button's name comes from its descendent elements by default. In this case, the `alt` attribute of the image provides the name.

Sara is good source for accessibility information but I think you've misinterpreted that post. If an icon button uses an `img`, making the image's `alt` the name of the button works well. If the button's icon comes from an inline SVG, background image, or icon font, I recommend Sara's first technique, visually hidden text within the `button`; names from text nodes are more robust than from attributes (`alt` being an exception, in part because it has been around far longer).

Thank you for the correction! Always good to learn.

From recent reports (based on their layoffs), "large" means around 1800 employees. I know they're not all developers, but I don't think labeling the mod buttons is beyond them.

It's much easier to label buttons when you work for a small company, because you know exactly who does what and exactly who is responsible for labelling the buttons and you know exactly what's going to happen to you when you label them, so you just do it and have that debate whenever it happens. Bystander effect, etc.

It's true but that falls more into the explanation-not-excuse category. I bet there's someone advocating for better ad partner integration and driving projects for that - no reason they can't have it for UX/UI/Accessibility too.

Eh, sounds unprofitable