Another thing to point out about XML's use of space is that, barring compression, the data is being stored as characters wrapped in tags, making storage of integer, numeric, or floating point data LESS efficient.

Datatype SQL Length XML XML Length

--------- ----------- ---- ------------

TinyInt 1 Byte 255 10

SmallInt 2 Bytes 32767 12

Int 4 Bytes 2147483647 17

FLOAT 8 Bytes -2.23.40E-308 20

In the author's example, he chose numeric(10,8) to store the rainfall amount, which would require only 9 bytes for SQL to store, but 18 for XML to store (90.12345678).

The author also chose to impose an [hour] column on the conventional table, which was unnecessary, because it could have been included as part of the [date] column.

Finally, he could have chosen to have a separate column for each hour in his schema for the conventional table, rather than making it one row per hour. (That wouldn't have been my first choice, because I feel it lacks flexibility in manipulating the data. But it is essentially the approach that was taken with the table that used XML.)