2007033 - Remove unused line-number object attribute

Status: NEW

(Core :: Disability Access APIs, defect)

Description

[James Teh [:Jamie]]

For LocalAccessible, we expose a line-number attribute with the caret line number. We never implemented this for RemoteAccessible. We should either:

1. Implement it for RemoteAccessible.
2. Remove the attribute altogether.
   • Neither NVDA nor Orca seem to use it. I'm not sure about JAWS or other clients. That said, no one has complained since we shipped Cache the World over 2 years ago, so maybe no one needs this on Windows or Linux. Chromium doesn't implement it either.
   • Chromium does implement AXInsertionPointLineNumber though, so we still need our implementation of that.
   • If we do remove this, we'll need to figure out how to test CaretLineNumber in browser tests, since that currently relies on the object attribute. The simplest option would be to add a method to XPCOM nsIAccessibleText.

Comment 1

[James Teh [:Jamie]]

Eitan, do you have an opinion here? Unshipping things always seems scary, but as noted in comment 0, no one seems to care. Especially given that Attributes is already expensive, it seems ideal to make it slightly less expensive by avoiding something no one apparently uses.

Comment 2

[James Teh [:Jamie]]

Oh interesting. Running strings on some relevant JAWS binaries (specifically FSDomNodeIAText.dll), I do see "line-number". So apparently they do use it for something, though I don't know whether they care, since Chromium doesn't implement it.

Comment 3

[Eitan Isaacson [:eeejay]]

Are you sure that the presence of "line-number" in that binary means they are looking at that attribute? I would remove this in either case since it has been years at this point where we didn't actually support this.

Comment 4

[James Teh [:Jamie]]

(In reply to Eitan Isaacson [:eeejay] from comment #3)

Are you sure that the presence of "line-number" in that binary means they are looking at that attribute?

I can't be certain, but IMO, it is very likely. That dll is entirely about IAccessibleText, so unless they use the string "line-number" internally for some reason related to their IA2 implementation (and I can't see why they would), it is very likely the object attribute.

Comment 5

[James Teh [:Jamie]]

Interestingly, NVDA does have an option to report line numbers, but it doesn't work in Firefox (even for LocalAccessible) because it's not implemented for IA2. It does work in Microsoft Word, for example.

Comment 6

[James Teh [:Jamie]]

I was slightly concerned that removing this would regress things for Thunderbird and JAWS, since we do expose line-number there because Thunderbird's composer is in the parent process (LocalAccessible). However, I couldn't find any setting nor command to report line numbers in JAWS.