Microsoft's OOXML Formulas Could Be Dangerous

hill101 writes "According to Rob Weir's blog, Microsoft's 325-page OOXML specification for spreadsheet formulas is deeply flawed. From basic trigonometric functions that forget to specify units, to statistical functions, to critical financial functions — the specification does not contain correct formulas that could possibly be implemented in an interoperable way. Quoting Mr. Weir: 'It has incorrect formulas that, if implemented according to the standard, may cause loss of life, property, and capital... Shame on all those who praised and continue to praise the OOXML formula specification without actually reading it.'"

Comments

Re:Surely we all saw this coming

[simong]

Depends what you call mission critical. Excel was used in voting systems in the Scottish and local elections in the UK this year, and Diebold's GEMS system uses Access as a database. High tech indeed.

Guess what?

[sid0]

Type in =SIN(30 degrees) if you want degrees. I'm sorry, Excel doesn't pander to high school students. In the real world, when the sine of an angle is mentioned, it is SUPPOSED to be radians. Every programming language I know accepts arguments for trig functions as radians.

The article, or at least this part, is FUD.

Re:Guess what?

[azrider]

Touche on the high school part, forgot about that. :) The main customers are of course still enterprises. I still think the default argument should be radians.

Who cares whether the default argument is degrees or radians. Two things are missing from the quoted "specification" document:

1: What is the default argument

2: Is the specification consistent across all functions which use this type of value as an argument

A specification which conforms to neither proper or common usage is worse than no specification at all

This is what Rob Weir was saying.

Re:Guess what?

[Hoi Polloi]

The number 360 isn't arbitrary. It stems from mesopotamia and the need to represent numbers as fractions. People needed a number that could be divided by 2, 3, 4, 10, 12, etc easily in one's head yet big enough to provide small fractions.

From wikipedia:

"The number 360 as the number of 'degrees' (i.e. smallest practical sub-arcs) in a circle, and hence the unit of a degree as a sub-arc of 1360 of the circle, was probably adopted because it approximates the number of days in a year. Its use is often said to originate from the methods of the ancient Babylonians. Ancient astronomers noticed that the stars in the sky, which circle the celestial pole every day, seem to advance in that circle by approximately one-360th of a circle, i.e. one degree, each day. Primitive calendars, such as the Persian Calendar used 360 days for a year. Its application to measuring angles in geometry can possibly be traced to Thales who popularized geometry among the Greeks and lived in Anatolia (modern western Turkey) among people who had dealings with Egypt and Babylon.

Another motivation for choosing the number 360 is that it is readily divisible: 360 has 24 divisors (including 1 and 360), including every number from 1 to 10 except 7. For the number of degrees in a circle to be divisible by every number from 1 to 10, there would need to be 2520 degrees in a circle, which is a much less convenient number.

Divisors of 360: 1, 2, 3, 4, 5, 6, 8, 9, 10, 12, 15, 18, 20, 24, 30, 36, 40, 45, 60, 72, 90, 120, 180, 360"

Re:MS Office approx. Reference Implementation

[Lonewolf666]

-Which version of MS Office (including level of Service pack)?
-And who guarantees that the "reference implementation" is still available 5 years from now? (hint: Microsoft tends to discontinue sale of its products after a few years).

MS claims this is a FULLY DEFINED STANDARD

[tinkerghost]

When other people claim a standard is fully defined, it means that all the standard use cases are defined* - units, expected parameters, optional parameters, etc. In the real world, nobody uses radians. Radians are used by engineers & scientists. Pilots, backyard builders, school children, and the occasional office worker use degrees.

To be honest, nobody cares if OOXML defines SIN(x) to take radians, degrees, gradians, or hyperbian-arc-vectors. What we care about is that someplace in the fully defined standard, OOXML needs to say:

DEFINE: SIN(x[,unit])

• SIN: geometric function dictating the height of a right triange with a hypotonous of 1 and an angle of x.
• x: parameter describing the angle to be operated on by the sin function
• unit - optional: one of a predescribed list of standard descriptors for angle:
  1. D: Degree - unit of angle defined as 1/360th of a full circle
  2. R: Radian - defined as the angle at which the length of an arc is the same as the radius of the arc. 1/2Pi of a full circle ~ 57.3 degrees
  3. G: Gradian - unit of angle defined as 1/400th of a full cicle.
  Missing unit parameters are defaulted to Radians. Unknown unit parameters will result in a type error.

That's how a proper standard useable for international work in multiple fields is defined. You do not just dump your US help file into the standard & call it done. I have had to deal with a lot of standards, both Military and Industrial, the OOXML standard is well below the grade of the average Mil or Ind standard.

That's before you get to the point of inclusions in the standard like "Must Replicate Office 98 Behaviour for this feature". Now, if there was a reference to another standard that defined Office 98 behaviour, then it's not a problem. However, I don't see a reference included in the OOXML standard. Worse, for dates, OOXML defines the proper behaviour as their broken implimentation of the Gregorian Calendar - a direct conflict to the existing ISO standards.

I don't care who sponsored this standard, it's not a properly writen standard. It has huge holes & it's contradictory to several existing standards. Either one should get it rejected. If MS cleans it up so it meets the actual requirements of a "STANDARD" then they should get approved. If they leave it as the crap heap it is, it should be rejected.

*- if passing sqr(-6) as a unit works in the implimentation, that's not the standards problem. However, if the standard fails to mention the default unit type & the existance of the unit parameter, then there's an issue.

Re:Impartial reviews

[armb]

> While I fully agree that the rules of English are screwed up, you need to put your trailing comma before the closing quote

That's the rule for American English. British English is often more logical.
http://grammar.ccc.commnet.edu/grammar/marks/quota tion.htm
http://www.catb.org/jargon/html/writing-style.html