Stewart very rightly draws attention to the practical value of the knot as a measure of velocity. In so doing he underlines one of the defects of SI: the essential simplicity and uniformity so attractive when considered in abstract becomes rather blurred when put into practice. The knot falls completely foul of SI; neither the unit of length nor that of time is an SI unit. The practical application of both units illustrate the difficulties well.

The choice of a standard set of prefixes to units, each separated by a factor of 1000, is both logical and convenient. One need never write down a number less than unity or greater than 999·999. … If the number is 1000 or more, it is divided by 1000 until it does lie within these limits and the appropriate prefix added to the unit. If it is less than unity, multiplication is substituted for division. For length measurements one starts with the metre and goes up to the kilometre, or down to the millimetre, micrometre and so on. Perhaps the only person whom this does not suit is the astronomer who is likely to prefer the light-year (approximately 10 metres, the index 16 being outside the range of prefixes) or the parsec (2·46 light-years).

The world, however, is not a linear one and difficulties start when one measures areas.