It was introduced in beginning of times in bug 151048 as part of ATK work. It got its first implementation in bug 312093. It got a current face in bug 445513. It doesn't have equivalent in IAccessible2 (http://accessibility.linuxfoundation.org/a11yspecs/ia2/api/AccessibleText.idl) or ATK (https://developer.gnome.org/atk/stable/AtkText.html#AtkTextBoundary). However ATK-based screen readers can obtain it if proper value instead the AtkTextBoundary is given. Implementation is a combination of GetTextAttributes and GetText, it doesn't seem very useful. On the another hand, all three methods getTextAt/Before/After shares same implementation what doesn't sound right. There's no real clarification why this constant was needed (except a short talk in bug 340809 comment #25) but still it doesn't seem I had a good understanding those times what this constants is for. So I suggest to remove it.

Comment on attachment 762021 [details] [diff] [review] patch >- case BOUNDARY_ATTRIBUTE_RANGE: >- { >- nsresult rv = GetTextAttributes(false, offset, >- aStartOffset, aEndOffset, nullptr); I fell like I saw some complexity in text attributes getting code that may have just been for this, it might be worth a look to see if you can rip more out :) r=me if you nuke the constant in nsIAccessibleText.idl too

(In reply to Trevor Saunders (:tbsaunde) from comment #2) > I fell like I saw some complexity in text attributes getting code that may > have just been for this, it might be worth a look to see if you can rip more > out :) no, no, the complexity is because of API > r=me if you nuke the constant in nsIAccessibleText.idl too yep, I forgot about it.