Alphasort

We solved the problem, for us, by starting every entry with an initial letter. So far, since upper or lower case can be of vital significance (e.g. MW, mW), we have had to accept the shortcoming inherent in the system, and have two alphasorts.

At the fastest reported sort rate of lGB per minute on the alphasort machine [NBC+ 94], re-computation of the cube for a data collection rate of 10GB a day would require roughly one day to sort 4 month data collection.

For example, in Spanish word lists the letter pairs ch and ll are normally sorted after c and l respectively. COCOA enables such a sorting order to be specified, so llano comes after luz rather than before it as would happen naturally with an English alphasort.