Is there a standardized method for accessibility and form semantics?

With so many articles on "correct", "semantic" and "accessible" use of forms and architecture, I am rethinking how I approach forms. There are literally so many variations on what is "correct" that I am not 100% more than accurate.

In an MDN article ( here ), he mentions:

In this example, the screen reader says "Small fruit juice size" for the first widget, "Medium fruit juice size" for the second, and "Fruit juice size large" for the third.

<form>
  <fieldset>
    <legend>Fruit juice size</legend>
    <p>
      <input type="radio" name="size" id="size_1" value="small" />
      <label for="size_1">Small</label>
    </p>
    <p>
      <input type="radio" name="size" id="size_2" value="medium" />
      <label for="size_2">Medium</label>
    </p>
    <p>
      <input type="radio" name="size" id="size_3" value="large" />
      <label for="size_3">Large</label>
    </p>
  </fieldset>
</form>

      

        
        
        
      

    

Now I can see the benefit of this for something like the above example, however, let's say I made a multi-step shopping cart, I would not want the assistive technologies to say "Checkout: cc-number", "Checkout: cc-date" using "checkout" in front of each label. Especially in cases where steps are indicated. This would be repetitive and sometimes confusing, I would guess ... But I have always grouped form sections within <fieldset>

. Now I am rethinking with fieldset

and legend

in general, but is it going against semantics now? What is the compromise? Is there a balance?

Also, and I will use the same MDN article so that I don't send you all over the internet,

Note that even disregarding assistive technology, having a formal set of labels for a given widget allows users to click on a shortcut to activate the corresponding widget in all browsers. This is especially useful for radio buttons and check boxes.

Some assistive technologies may have problems handling multiple shortcuts for a single widget. Because of this, you must insert a widget inside its corresponding element to create an accessible form.

<form>
  <p>
    <input type="checkbox" id="taste_1" name="taste_cherry" value="1">
    <label for="taste_1">I like cherry</label>
  </p>
  <p>
    <label for="taste_2">
      <input type="checkbox" id="taste_2" name="taste_banana" value="1">
      I like banana
    </label>
  </p>
</form>

      

        
        
        
      

    

Now in this case, labeling for both of these elements is fairly common, I've used both methods, but is there a balance between accessibility and semantics here? I tend to put the label without wrapping the input for code consistency, and I know their strong discussion on this (basically the ability to opt out for

and not need id

and / or have labels in different areas of the markup). So, I'm not trying to rephrase the debate here, I tend to use for

and id

, regardless of whether I wrap the elements in label

or not. But if there is an accessibility issue for one, then why is the latter not the gold standard?

One more point, WAI-Aria rules are now contributing, so how much do these rules really affect the accessibility and semantics of the form?

<p>Required fields are followed by <strong><abbr title="required">*</abbr></strong>.</p>
<form>
  <fieldset>
    <legend>Fruit juice size*</legend>
    <p>
      <label for="size_1">
        <input type="radio" name="size" id="size_1" value="small" aria-labelledby="size_1_label" />
        <span id="size_1_label" required aria-required="true">Small</span>
      </label>
    </p>
    <p>
      <label for="size_2">
        <input type="radio" name="size" id="size_2" value="medium" aria-labelledby="size_2_label" />
        <span id="size_2_label">Medium</span>
      </label>
    </p>
    <p>
      <label for="size_3">
        <input type="radio" name="size" id="size_3" value="large" aria-labelledby="size_3_label" />
        <span id="size_3_label">Large</span>
      </label>
    </p>
  </fieldset>
</form>

      

        
        
        
      

    

I'm really just wondering if there is a standardized way to approach when dealing with semantic and accessible markup. So far, it seems like people are just doing what they think is right and then vocalizing their thoughts all over the internet.

I have read the W3C drafts and guidelines, but even they use different examples. Does anyone have any evidence of an approach that actually improves accessibility and semantics with respect to forms? Are there any specific websites that have time to check this out and come up with an accurate conclusion that I could consider?