List style resolution

[Svante Schubert <svante.schubert@gmail.com>]

The resolution of list styles is not easy to understand in the ODF 1.2 specification. I have just written a mail to the ODF TC to suggest on an improvement (see https://lists.oasis-open.org/archives/office/201302/msg00000.html).

For us the interoperability of ODF applications using lists is of interest.
I created a test document that I would like to share with the OIC TC and suggest to discuss on the next ODF Plugfest.
https://dl.dropbox.com/u/49473263/ListStyleResolution.odt

The test document is a valid ODF document that I have manually adapted to test all possible style references for a one and two level list - this could be done for up to 10 levels, but the change to find a new behavior is low and I try to avoid such huge efforts in my spare time. ;)

In the case when all possible references are set the following is applied:

1. The highest list references to a style of upper case roman numbering, its list item to upper case alphabet numbering.
2. The second list references to a style of lower case roman numbering, its list item to lower case alphabet numbering.
3. The paragraph style references to an image list with red bullets and should be triggered, when no list style is given elsewhere.
   

If you open the test document in your favorite ODF application, you will realize list labels within the test, that are not mentioned above.
The reason is that ODF defines the implementation defined behavior, when a list has no style at all.
What ODF user using multiple ODF application could ever desire such a thing?
Currently Calligra uses a numbering list, MSO15 uses no list and latest AOO/LO uses bullet lists.
Even if not every application would claim that their behavior should become the new default behavior, the two other applications need to differentiate old ODF document (where an old default list had being used) and new ODF versions, where a default list might be defined.
This can only work, by a differentiator like the version attribute.

Another topic would be how to minimize the problem resulting from an ODF application that writes out wrong ODF. For instance, if one ODF application within a working process involving multiple ODF applications has switched the colors red and green. How can this damage be undone?

PS: There has been a preliminary draft of the tests document above, that might be of interest as well:
https://dl.dropbox.com/u/49473263/ListStyleResolution-PreliminaryDraft.odt
This document only had a few possibilities of list style references, before I realized I needed to test all combinations for at least two levels. But in addition it tested @text:start-vale and @text:continue-list and showed interesting layout behavior in some ODF applications, when there is a heading in a list, a heading with not style, but only outline level and there are two paragraphs/headings in a single list item. This latter ODF scenario is useful, when there are two list items in a document each with a paragraph using distinct styles for instance first using green, second a red background. If the user has the cursor in front of the label of the red and press backspace, the label vanishes. When being saved, ODF now allows to move both paragraphs into a single list, which is for instance not allowed by OOXML.

PPS: The whole testing was triggered by some discussion on our last change-tracking call, where we realized that headings, lists and paragraphs are just paragraphs with different properties. Which is useful as they are often transformed to one another. For full discussion log, see https://lists.oasis-open.org/archives/office-collab/201301/msg00013.html

Best regards,
Svante