Explore landmark options in rendered notebook

[isabela-pf]

Problem and context

This issue comes from our user testing round 1: navigation. While completing tasks within the rendered notebook that required information from multiple cells and/or moving from one to another, it was mentioned that headings are helpful for navigating content breaks (@isabela-pf note: that is if notebook authors provide them and use them properly).

However, content breaks (ie. a new topic of discussion, the next step in a tutorial) are not the only structure in a notebook. It was mentioned that landmarks might be something that the participant(s) would find missing to navigate the non-narrative structure of a notebook such as page navigation or cells.

We need to investigate landmark options, common mistakes, and whether or not they are useful for the rendered notebook.

Possible solutions

At this point, the problem requires exploration. If we agree to fix this in a specific way, we can create and link an issue with implementation tasks. The solution will be to identify what path we want to take forward with landmarks.

As landmarks are ARIA, we also need to ensure that the landmarks are truly needed and helpful. They are not to be used as a bandage over poor HTML.

Acceptance criteria

This issue can be closed when we decide

• What rendered notebook structures are missing navigation based on further feedback
• If problem needs ARIA landmarks to have a better experience or can be solved by HTML improvements
• Where landmarks would fit within the ecosystem (is this a conversion step, is it a problem in the file format, is it in the hosting platform, etc.?)
• What next steps we want to take with landmarks on this project

Tasks to complete

Tasks are to be determined by further tests and team discussion.

[smythp]

✨

Here is your sparkles / magic star emoji as discussed in the meeting.

[smythp]

This is probably related to #19 , which suggests adding sections for each cell. This would create a landmark for every cell.

[isabela-pf]

I'm following up on this as I clean up our content type test results.

In our second round of tests, the structural changes we experimented with to make cell divisions more clear did not surface clearly to participants. Markdown cells read as text all at once, including if there are multiple Markdown cells in a row. Code cells have some visual distinctions with their borders, shading, and execution number. This round of tests did make execution numbers easier for most participants to find.

In further discussion on the results with @Jenneh, I would also like to note that this didn't usually stop participants from being able to complete a task. It was most likely to slow participants down or create some confusion. That isn't ideal. Still, we may be finding this issue in higher rates in user tests because we are asking participants to navigate not just from top to bottom of a notebook. We may be biased in this situation.

This doesn't answer the questions posed above, but I am mentioning it so we are aware this issue arose in the second round of tests as well. 🌻