Submitter's message Greetings!
The draft minutes for the ODF TC teleconference on 28 September 2026 are attached and/or pasted below.
Hope everyone is at the start of a great week!
Patrick
************Minutes***************
Patrick: Francis, Patrick, Regina, Svante, Olivier, Michael - quorum
Patrick: Minutes of Sept. 21st - approved by consent
Patrick: 3. action items no progress
Patrick: # Svante/Francis WG4 meeting for ISO
Patrick: # Question of how to approach deriving formulas - math
Patrick: With those additions - agenda is approved
Patrick: Svante/Francis - WG4 meeting - meeting with experts - take away - WG4 has more of a problem than we do, are wanting to address the fact OOXML doesn't specify functions that are in ODF. What about other functions added to OOXML or ODF but not documented in other standards - look at functions that are in both sets - are they defined the same way? if not, why not? - functions that aren't in either one -
Patrick: Svante - we need to argue this is more important than other projects - for MS - Google needs earlier notice of the meetings - Svante - test functions are easy to create - has Claude function to compare spreadsheets - need implementors to do this work -
Svante Schubert: The meeting focused on improving spreadsheet-function interoperability between ODF/OpenFormula, OOXML, and Excels extensions, with interest in including Google Sheets. Participants supported further investigation, but no new standardization project or deliverable was agreed.
Main points:
- Documentation gaps: Murata compared five functionsACOT, ACOTH, BINOM.DIST.RANGE, IFNA, and NUMBERVALUE. For these examples, ODF provided semantic definitions, while Microsofts extension specification mainly supplied syntax. Excel Help added behavioral details but sometimes contained ambiguities.
- A concrete difference: ODFs NUMBERVALUE requires a single-character decimal separator; Excel Help says that a longer separator uses only its first character. Participants debated whether standards should specify such error handling or leave it to implementations. This remained unresolved.
- Compatibility is more complicated than matching names: User-visible names can differ from names stored in files. Namespaces, legacy variants, and historical behavior must be considered to avoid breaking existing documents.
- ODF work is ongoing: Additional functions are being prepared for future ODF versions. Issue numbers in the comparison spreadsheet indicate work in progress, not necessarily mature or approved definitions.
- Testing could support the work: You advocated comparing actual results across applications and using test suites alongside specifications. Existing OpenFormula tests, LibreOffice tests, and a community comparison of Excel, Google Sheets, and LibreOffice were identified as potential starting points.
- Resources and scope matter: Microsoft could consider a narrowly defined request for expertise or tests, but made no resource commitment. Participants suggested reviewing manageable groups of functions.
- Publication options remain open: Possibilities included a technical report, changes to OOXML, references to ODF, or a jointly developed document. Permission to reuse ODF text would need clarification.
Follow-up commitments and requests:
- Murata: Publish his two discussion documents as WG4 documents and share them with the ODF TC.
- Francis: Forward Patricks detailed 2009 OOXML/OpenFormula comparison, with permission for wider circulation.
- John and Francis: Circulate draft minutes to participants for comments before publication.
- Participants: Submit written proposals before the next meeting, including suggested scope and priorities. No detailed technical assignments were made.
The next meeting was described as October 12/13, depending on time zone, although the transcript is unclear on the exact scheduling details.
Patrick: ## How to approach formulas - math?
Svante Schubert: A project showing how major implementations handle spreadsheet functions:
https://canispreadsheet.com/ and its code at
https://github.com/xxaflabsxx/spreadsheet-compat/
ODF 1.2 used a test case generator (examples included):
https://github.com/oasis-tcs/odf-
tc/tree/master/src/test/resources/odf1.2/tools/formulas-testsuite-generator
LibreOffice test files exist at
https://git.libreoffice.org/core/+/refs/heads/master/sc/qa/unit/data/functions.
Svante Schubert:
https://github.com/oasis-tcs/odf-tc/tree/master/src/test/resources/odf1.2/tools/formulas-testsuite-generator
Regina Henschel:
https://opengrok.libreoffice.org/xref/core/sc/qa/unit/data/functions/
Svante Schubert: And the D for download is here:
https://opengrok.libreoffice.org/xref/core/sc/qa/unit/data/functions/addin/fods/
Svante Schubert: Regina states, these are test documents that are used in Unit test in LibreOffice for function tests with expected results
Patrick: Regina - test documents for ODF 1.2 - implementation defined - fails. + operator for complex numbers
Patrick: One application may return an error, another a value
Patrick: Regina - specification does this already - inverse for example - we don't say how it must be done
Regina Henschel: Datedif
Patrick: Regina - there are functions that are incomplete
Patrick: Ah, thanks!
Patrick: Svante - interoperability between gnumeric, ooxml, google, lo
Patrick: Francis - interoperability - certainly a goal ; if given the standard, would an interoperability be possible
Patrick: Regina - need not have the same result - interation might give a different result - as a start value
Patrick: Olivier - approximations - can give a set of data inputs and outputs, and sensitivity of results - also called tolerance -
Patrick: Svante - conformance
Patrick: Regina - has examples in function definitions - path towards documentation
Patrick: discussion of ISO access issues
Patrick: Francis - examples are non-normative -
Patrick: Svante - JPEG has code in the standard - thinks test suite is important
Patrick: Regina - comments in early issue - example were not included - can go with a test suite - for example the CSS standard - they have a test suite and see if your browser is correct or not. - would be good to have for functions.
Patrick: Francis - examples only cover a handful of cases - can't claim conformance with just those examples - having a formula has an advantage - if you don't meet the formula - then you don't conform
Patrick: Francis Absolute function - not a defining behavior
Patrick: Francis - indicative of possible conformance
Patrick: Regina - we have nearly 700 functions -
Patrick: Svante - we need to use AI
Patrick: Regina - need to have a public github issue -
Patrick: Svante - Action item to create such an issue
Patrick: Francis - starting with functions already in ODF and OOXML - 1. are implementation interoperable 2. are the specifications equivalent
Patrick: Francis - two parallel activities
Patrick: Svante - black box - same data in, same result
Patrick: Svante - compare same name, similar names, functions in one set and not the other. - start with same name.
Patrick: Francis - should not allow this to divert us from preparation of ODF and the next version.
Patrick: Adjourned
-- Patrick Durusau
---------------------------------
Patrick Durusau
Patrick Durusau (Personal)
---------------------------------