Talk:Storing and processing imprecise data

Thoughts

Suppose that we conceive of a compact form that contains only the date/time information and its uncertainty properties. I can also imagine a full-detail format that contains references justifying the date/time information and/or the uncertainty properties.

There should probably be a compact binary form suitable for a fixed-length database column.

There should probably be a a compact and an full-detail XML format.

Do we also need a full-detail binary format? Assume no for now.

Dmclean 04:04, 4 November 2007 (UTC)

Thoughts, part 2

Let us say that the data starts with two tokens. One that defines the granularity of the value, and one that defines the precision properties of it.

Granularity could vary anywhere between fractions of a second to billions of years.

There are three categories of precision: the value given is precisely known within the given granularity; the value given has parts with varying levels of precision (to be followed by separate tokens for the various relevant levels of granularity); the value given is imprecise and a probability function is given.

It may be good to have a token for the date system. Date components may be precisely known in one system that do not map to precisely known values in other systems (such as the Hebrew calendar).

Dmclean 04:04, 4 November 2007 (UTC)

Thoughts, part 3

Let's suppose that rather than require a series of tokens to specify anything that we start with a code token, that can specify a number of commonly used modes but has a "specify everything" mode.

Dmclean 04:04, 4 November 2007 (UTC)