Re: [office-formula] TC review comments, hiding test cases (remove fromformal spec)

This sounds good.  Thanks, David.
 

But let's continue pushing OASIS and
ISO on this  as well.  Although short term there is not much
we can do, longer term I think we can turn them around.  

-Rob

___________________________

Rob Weir

Software Architect

Workplace, Portal and Collaboration Software

IBM Software Group

email: 

phone: 1-978-399-7122

blog: http://www.robweir.com/blog/

"David A. Wheeler" <>
wrote on 05/11/2007 12:55:42 AM:

> As you know, we (the formula subcommittee) have submitted our draft

> to the TC for commentary, and been awaiting their feedback.  Here

> are the comments I've received so far, and what I think we need to

> do.. but if you disagree, please speak up ASAP.

> 

> 1. Test cases.

> 

> The main discussions have been on objections to including test cases

> in the spec, particularly as being normative.  We've been going
back

> and forth on this issue, and frankly, I remain quite unconvinced.
 

> One argument is that all we need to do is write perfect, unambiguous

> text without using test cases. That's a goal I laud and we should

> strive towards.. but since no previous spec has achieved that, doing

> the same thing as previous efforts but expecting a different result

> seems closer to the definition of insanity.  (I'm alluding to
an old

> joke, "Insanity is when you do the same thing repeatedly but
expect 

> something different to happen.")  Another argument is that
test 

> cases can't cover all cases - true, but if their purpose is to 

> clarify the text and ensure correct interpretation, that's a non-

> issue.  Antoher argument is that even including tests cases there

> may be ambiguity; which is true, but if they eliminate 80% of them,

> then that is still a good thing.  All agree that test cases are

> valuable and important, but yet somehow they're not important enough

> to be in a spec (!).

> 

> But our goal is to produce a spec, not an argument.  It appears
very

> likely that including this information, even though it's very 

> useful, will IMPEDE its becoming a standard. So even though I think

> this is NOT a good idea, I think we must NOT include the test cases

> in the official spec so the spec can be accepted. Instead, they 

> should be in hidden text along with the other commentary. We might

> include 1-2 examples in each, which would help slightly.  This

> change will require some rewriting of text since what was 

> unambiguous before might become ambiguous.

> 

> The test cases aren't being destroyed; they'll still be included in

> the annotated document, and of course they'll still be useful in 

> determining what the specification SHOULD say.

> 

> 

> 2. Hiding Commentary

> 

> The first page clearly states that the notes, etc., will not be in

> the specification, but many reviewers just want to see what the 

> FINAL spec will look like, and DON'T want to see all the rationale

> for why it is the way it is.  So we need to hide the other 

> commentary, and turn on hiding by default.  Those who want to
see 

> the commentary can just turn it back on.

> 

> This was always the plan; I'd delayed it because we were going to

> switch formats, and that seemed like the best time to do it.  But

> sounds like we need to do it a little sooner.

> 

> 

> 3. Basis/YEARFRAC definition

> 

> We need to define the Basis/YEARFRAC stuff more exactly (esp. with

> the test cases removed).   This is a specific technical comment

> which I agree with.

> 

> 

> I intend to walk through the document and hide all the commentary

> and test cases, and then repost.

> 

> Sound okay?  Let me know of any issues.  Thanks.

> 

> --- David A. Wheeler