Tables without captions are less accessible for people with screen readers (and often less informative to everyone). Setting the default property to 'on' would probably increase the number of people who add captions.
Many screen readers have the ability to call up a list of table captions and allow the user to jump directly to a table, which is very helpful for users who wish to revisit a table, or just look up information from a particular table. Where a table immediately follows a section heading, the caption becomes somewhat redundant if it simply repeats the heading, because screen readers can generally navigate to particular sections (and many users will be accustomed to navigating Wikipedia articles in that manner).
That means that we shouldn't be forcing captions to be written mechanistically, but nevertheless users of screen readers will tell you that adding captions to table is one of the most helpful improvements to their experience that we can make. Setting the caption to "on" by default won't make it any harder for editors to omit a redundant caption, but should encourage editors who would not otherwise consider providing a useful caption to do so.
I tried the effects of these two changes. Even if I totally support the idea presented, I must say I do have some mild concerns with the way this is currently presented to users of the VisualEditor. I believe it's possible to resolve this quite easily:
- In all cases where users decide to ignore the empty space above the table they are going to insert, a super-cryptic |+ ends in the wikitext. I think this is effect is unavoidable, but:
- VisualEditor presents the caption as an empty space. There is absolutely zero hint at what this space is meant to be. If I focus it, it gets a thin dashed line and an editor cursor. But that's it.
- Simply repeat the word "Caption" as a bright gray placeholder text.
- When I focus the caption, the popup that allows me to open the tables preferences (and disable the caption) must show up. Currently it doesn't.
The |+ is of course the wiki-markup for a caption, so there's nothing more cryptic about it than any other markup.
I'm not sure who the audience might be that would find any problems with the change.
If an editor is accustomed to wiki-text, there's nothing new about |+ and they will know the caption is blank. If the editor normally uses VE, then they see a empty space where captions normally go, so they can see the table has an empty caption. In either case, another editor may be encouraged to add a missing caption.
I suppose an editor who is not used to reading our articles might not expect that we put captions immediately above our tables - is that it?
I agree that the focusing is inconsistent, though. Putting the focus anywhere on the table body initially opens the properties pop-up; whereas putting the focus on the table caption does not.
[…] there's nothing more cryptic about it than any other markup.
I agree. This was only meant as a note.
If the editor […] uses VE […] they can see the table has an empty caption.
This is exactly where I disagree. Yes, I can see there is some empty space. But how should an unexperienced user know what this empty space is supposed to hold? Since this ticket here is about teaching unexperienced users, I suggest to do something about this. I outlined a very trivial solution. It's also possibly to test this very easily with real-world people, which I suggest to do.
Yes - it occurred to me that a placeholder would be ideal, but as we don't have any mechanisms for sub-document placeholders, would require a but more work.
In all cases where users decide to ignore the empty space above the table they are going to insert, a super-cryptic |+ ends in the wikitext. I think this is effect is unavoidable, but:
This isn't unavoidable - we mark some paragraphs as generated-empty, and on conversion to HTML discard them if they were left empty by the user, so we could do something similar with table captions. Again it would require a fair bit more work.