@jashkenas/text > @observable/text

This Inputs / Jeremy Ashkenas / Observable


Is more beautiful than this Observable Input: Text / Observable / Observable


Why the former is not the default? It will add more beauty.

The assessment of one as “more beautiful” than the other is clearly rather subjective. The new version is definitely superior from the viewpoint of Web Accessibility.

Of course, there are things you can do to style the output. We could make the label text bold and larger, while making it take up a bit less space as well:

1 Like

Observable notebook are so dynamic, that I can not imagine how people with disabilities can work with them.

I agree that design is subjective, and the only way to reach definitive conclusion on that is to use A/B testing. What I find frustrating is that alternative input designs are not getting as much exposure if we would try to compare the usage. I personally just wanted to repeat the form from Module require debugger / Observable / Observable and took me a while to notice that it is not a standard Inputs.text.

The point is that the output produced by Observable should be accessible. If you want to design things that people use, then you should at least give this some degree of consideration.

Do you mean that Inputs / Jeremy Ashkenas / Observable can not be made accessible without sacrificing its visual design?

Yes, that’s fundamentally correct. The major issue is that the labels generated by new Inputs object are proper HTML label elements. This allows assistive screen readers to properly associate the label with the input element. In Jeremy’s (admittedly, very nice looking) version the label is just a div. :frowning:

Of course, I had nothing to do with the creation of the new Inputs but I’d be surprised if this was not a major design consideration.

I think the markup can be changed to use <label> + CSS without sacrificing the design. But just using recommended elements is not enough to prove that Observable is accessible. At least it not convincing for me if there is no real test with NVDA and target users had been done, and probably not just a single test, but testing on a regular basis.