Summary 2010-09-14 of OpenFormula meeting

["David A. Wheeler" <dwheeler@dwheeler.com>]

Summary 2010-09-14 of OpenFormula meeting

(As always, please reply-all with corrections.)

Attendees:
David A. Wheeler
Andreas G.
Eric Patterson
Patrick D.
Dennis Hamilton
Rob Weir


Not attending:
Eike Rathke (he sent his regrets)


We WILL meet next Tuesday.

Note: For our current status, see the dashboard:
http://tools.oasis-open.org/issues/secure/Dashboard.jspa?selectPageId=10056



TOPICS:


Focused first on NEEDS-DISCUSSION, then Unresolved untargeted unassigned.

* OFFICE-3342
http://tools.oasis-open.org/issues/browse/OFFICE-3342

Dennis found some important stuff, pointed to by Andreas & others.
Dennis confirmed that we ONLY need to talk about IRIs.
Dennis has promised to provide the supporting information to document this
(e.g., document names & section names) in this JIRA comment.

Dennis will subdivide this for part 1, part 2, part 3
(part 3 is probably most challenging), and will remove NEEDS-DISCUSSION.

So there will be an update within 48 hours on this one.
For part 2, URI will change to IRI everywhere (essentially).
IRI includes URI as a subset.

This will shorten syntax definition considerably, makes it crisper.


* OFFICE-3040
http://tools.oasis-open.org/issues/browse/OFFICE-3040

Rob has verified that this is the correct formula.

Eric P: This is not a function that Excel currently has.

Andreas: OOo does use the corrected formula.

Rob will remove NEEDS-DISCUSSION, and state this resolution.


* OFFICE-2663 (What is a "character"?)
http://tools.oasis-open.org/issues/browse/OFFICE-2663


Dennis: The resulting value may differ depending on normalization.

Rob: We're saying whether or not normalization occurs;
either it does or it doesn't.

Wheeler: We're allowing normalization, but not requiring it.

Wheeler: I believe *all* functions that accept/product Text types
are affected by normalization.

{Long complication discussion here, below is only a subset.}

Dennis: I'd like to say that operations just operate on characters.

Rob: Then we're back to "what is a character".

Patrick: Alex's original comment was about the CODE function.
As far as the original comment is concerned, so we've literally
answered this comment, and normalization is not an issue.

Rob: We found a broader issue.

Wheeler: The UNICODE function does have the broader issue.

Dennis: XPATH is very careful to avoid this issue.

Rob: That's equivalent to what I say as "undefined".

Wheeler: What about combining characters?

Rob: Think about LEN; is char followed by combining char one character?

Rob: The output could be in a different encoding.

Wheeler: Imagine concatenating with a cell value, and that
cell contains just a combining character.

Dennis: Right.

Rob: We need Unicode normalization functions in ODF-Next.

Wheeler: This clearly NEEDS-DISCUSSION for the future.

Dennis: I doubt implementations' functions normalize text on every input.
Systems may normalize when they read from users
{or when reading a file?}.

Rob: With Thai, etc., they will get composed.
Another good read, but only a draft, is a W3C note on normalization.

Eric: Aren't we overthinking this?

Dennis: Simply just not caring is easy.  E.G., letter 'e' is a char,
a combining char is a char, concatenating them is length 2.
If we ran it through normalization then we'd get length 1, but
that's a different issue.

Eric: We aren't defining what happens from file->memory, or from
input->memory.  We just need to say what happens when they get what they get.

Rob: Could say "no additional normalization applied".

Wheeler: "It is implementation-defined if normalization is applied while
reading text or from the user."

Wheeler: You can't know if it's normalized.

(Will continue to discuss.)



BELOW: Unresolved untargeted unassigned comments.


* http://tools.oasis-open.org/issues/browse/OFFICE-2702

ODF-Next.

* http://tools.oasis-open.org/issues/browse/OFFICE-3426

No action.   That capability exists, we have a FORMULA function.


* http://tools.oasis-open.org/issues/browse/OFFICE-3428

Dennis: I've done this manually by saving as CSV.

Andreas: Paste special values can replace formulas with values (OOo, Gnumeric, Excel).

Rob: Nobody has multiple formula engines, so this is a non-problem.

Dennis: There are kind-of ways to do it, though not as simple as throwing a switch.

This is late, and we know of no spreadsheet implementations that do this directly,
so it seems unwise to add this to the specification.

Eric: There are some add-on packages that do this sort of thing, e.g.,
Monte Carlo simulators.

We'll put this in ODF-Next.

We WILL meet next week.

--- David A. Wheeler