The discussion regarding the Mars Climate Orbiter is not completely correct.
> The error was not at the final destination, as the paragraph seems to imply, but throughout the mission, since that was part of the course correction process.
> The IEEE's investigation revealed a systemic failure of engineering oversight and a generally complacent response to anomalous results.
> There was a major failure in interface control in the design phase, as that error should have been obvious in the interface control specifications that were part of NASA's systems engineering process. It was not detected during design reviews and apparently never tested during system integration and checkout.
What should have been detected within the first couple of times where the craft was already off-course, well before getting to Mars, was completely ignored.
We need methods to distinguish between items we consider independent (i.e. of independent kind)
We need that independence to be checked automatically
We need to be able to require as quantity to be of a certain kind
We need results conversion to indicate when we have the kind wrong.
I support the need for some mechanism for asserting (checking) that a parameter if of a required "kind"
(e.g. MustBe(x,Qty) returns x, or error if UnitsOf(x/Qty)<>1 is not dimensionless (SIUnitsOf in newer versions, but I mean all the dimensions available in MathCAD)
I would support a results check that warns if the units scaling in the result place holder does not create a unitless numeric value,
i.e. when MathCAD determines there is a missing conversion factor (e.g. a missing "/kg") it shows those factors in red. I wouldn't go as far (yet) as to suggest that a blank placeholder creates a warning.
It is my view that the 'dimensions' concept be extended to provide more spare user dimensions (of base value 1) so that users can tag values and quantities that are of a different 'kind'.
This would be a multiplicative supplement, such as Kg.DrySteam and Kg.WetSteam where DrySteam and WetSteam are the names assigned (by the user in the worksheet) to the two extra 'dimensions' they used to identify these independent kinds of quantity. So we could not then add kg + Kg.DrySteam + Kg.WetSteam as the dimensions would not match [M+M.U0+M.U1] (U0 and U1 being those extra dimension symbols)
Such symbols can be used for all the different types of checking, such as apples+oranges.
It may be that it (my suggested method) is simply a subtle difference in how we feel that MathCAD should mechanise the process of distinguishing between different kinds of things that we, as users desire to be kept distinct (i.e. can't be added).
MathCAD should enable the user to distinguish between "kinds of Quantities". I believe that can be done by including more dimensions, some of which are user named. There needs to be at least another 7 beyond the SI_&, Money, and Angle, i.e. 16 minimum (I chose that because it is one of those nice binary word sizes, and I tend not to expect that many distinctions being needed, but you may have a greater list of need. Can you justify 32?)
