Re: Reflow (original question)

Dear Guy and WAI Interest Group,



This does raise an interesting question. The text embedded in an image is
an image of text, clearly, and it may be important. It must be represented
by alt-text.

Phrasing this criterion was very difficult. The multimedia nature of web
content was a challenge. The primary goal of the criteria was to enable
reading of the text portion of a page. The goal was to eliminate two
dimensional scrolling to read text. Text is a semantically linear object
that is wrapped to fit the width of a viewport. Scrolling perpendicular to
the direction of reading is a tractable option for a person, but scrolling
in the direction of reading (the direction of the lines) causes gross
amounts of scrolling. If all languages arranged their lines in the same
direction we wouldn't need such cagy language. So we used the term
2-dimensional scrolling rather than assuming horizontal or vertical
scrolling.

Images are very easy to understand when you zoom them. You just start at
one corner and travel your way through. At 400% you will scroll about 16
times to see everything in an image that filled the screen originally. With
text you would probably have to scroll 50 to 100 times to comprehend the
same area. That is the problem. A 1000-page book text takes me about
100,000 scrolls to read. A thousand-page picture book would take me 16,000
scrolls. That's a lot, but even a typical reader would require 1000 scrolls.

You see the complexity of images for the reader in terms of physical
movements is not as critical.

The central problem for reflow is text and two-dimensional scrolling of
text. Images are another problem.

The reflow criterion was developed because enlargement prior to S.C. 1.4.10
was a mater of zooming text like it was one big rectangular picture of
text. The fact that typical readers though of traversing text like one
would traverse an image created a situation where these readers
underestimated the burden caused by reading text as a zoomed image. It was
an understandable mistake, but it caused havoc in the lives of readers with
low vision.

Best, Wayne



PS. I didn’t just pull these numbers. I analyzed the complexity of reading
with 2-dimensional scrolling rigorously.
This does raise an interesting question. The text embedded in an image is
an image of text, clearly, and it may be important. It must be represented
by alt-text.
Phrasing this criterion was very difficult. The multimedia nature of web
content was a challenge. The primary goal of the criteria was to enable
reading of the text portion of a page. The goal was to eliminate two
dimensional scrolling to read text. Text is a semantically linear object
that is wrapped to fit the width of a viewport. Scrolling perpendicular is
a tractable option for a person, but scrolling in the direction of reading
(of the lines) causes gross amounts of scrolling. If all languages arranged
their lines in the same direction we wouldn't need such cagy language. So
we used the term rather than assuming horizontal or vertical scrolling.
Images are very easy to understand when you zoom them. You just start at
one corner and travel your way through. At 400% you will scroll about 16
times to see everything in an image that filled the screen originally. With
text you would probably have to scroll 50 to 100 times to comprehend the
same area. That is the problem. A 1000 page book takes me about 100,000
scrolls to read. A thousand page picture would take me 16,000 scrolls.
That's a lot, but even a typical reader would require 1000 scrolls.
You see the complexity of images for the reader in terms of physical
movements is not as critical with images.
The central problem for reflow is text and two-dimensional scrolling of
text.

On Wed, Dec 27, 2023 at 4:33 PM Guy Hickling <guy.hickling@gmail.com> wrote:

> This is to answer Sarah's original query in the Reflow thread, and give my
> take on it (others may differ). Since that thread is now all about AI, I'm
> starting a new thread for this answer. It will probably help if any more
> comments on AI are kept in the earlier thread "Re: Reflow".
>
> First, this SC has an advisory technique specifically for images,
> Technique C37:"Using CSS max-width and height to fit images". It basically
> says that to avoid the need for a horizontal scroll bar at high zoom, we
> can limit the size of the image to ensure it stays within the available
> window or other container without scrolling - i.e. it doesn't zoom with the
> rest of the content. I think that recommendation partially answers this
> question - some images don't need to zoom to high levels. Those images
> should, I believe, be treated as covered by the SC and not zoom to the
> point of needing a scroll bar.
>
> For most mages, like the example you quote of photos in a news article,
> landscapes and townscapes for instance, or head shots of a person, that's
> probably ok, people can still understand the picture, albeit at a smaller
> size than if the image were zoomed with the other content. (On the other
> hand, a particular image might show an important small object somewhere in
> the scene, that would be lost to low vision users without zooming - that
> could be an exception.)
>
> But the SC specifically excepts mentions "images required for
> understanding (such as maps and diagrams)". Maps and charts and graphs
> usually have a lot of writing on them. To someone needing 400% or more zoom
> to read the surrounding text content, keeping a graph image small, aka
> Technique C37, would leave the text on the graph too small for them to
> read. Hence why the SC classifies that kind of image as an exception to the
> SC, and it's ok to zoom them beyond the viewport. The same would apply to
> an infographic with lots of text in it; they usually fail SC1.4.5 "Images
> of Text", of course, but if there was one that validly passed 1.4.5, then
> it would be an "image required for understanding" that could be allowed to
> expand beyond the viewport.
>
> So, critically, whether an image is caught by the Reflow SC, or can be
> excepted and allowed to zoom with a scroll bar, I think we have to decide
> whether, and by how much, the image is an "image required for
> understanding". That is often a subjective decision. Consider a head shot
> of the prime minister, for instance, in a news report. Is that
> informational? Not really, I would say, most people know what he looks like
> and such a photo would probably just be there to add visual interest to the
> report. But if the police published a similar head shot of a dangerous
> criminal with a warning to people to keep clear of them (no, not the prime
> minister!), that photo is highly important information so should be zoomed
> along with the other content for people with low vision needing high zoom.
> I hope this answers your query, and sorry for the delay in replying to your
> second query.
>
> BTW: I think the phrase used in the this SC and it's second Note,
> "two-dimensional layout", is somewhat confusing here, and hard to
> understand (and maybe why this question was asked). All images are by their
> nature two-dimensional, but I think only some images are in scope of the
> SC. You can't really say that a photo has "two-dimensional layout" in quite
> the same way a data table has. No one laid it out, it's just what the
> camera saw. A bit of clarification or re-wording to that Note might not
> come amiss.
>
>
>