When designing a specialized structured-data document format (perhaps upon XML): part of the requirements for this document format are that it accommodates, in a metadata section, a history of meaningful (app-level) changes to the structured data at a field level.
At minimum, useful tracked information would be:
- an author identifier
- time stamp
- type of change
- what it was changed from
Both data items and any lists of such data items are to be tracked meaningfully, efficiently. The data schema should be separable/unaware of the metadata tracking it – although facilitating annotations such as node identifiers could be required. A trusted application could be required to enforce the tracking; however, it would be a benefit to be able to calculate the “deltas” at intervals by comparing data-sections between versions rather than requiring the editor to track each change live.
“Meaningful” tracking may involve the metadata schema treating higher-level data changes atomically – such as an update to a group of fields which is treated at the application level as one data-point.
For character-by-character or byte-by-byte data, diff/patch type algorithms work. Structured data (to be treated as structured) seems to me to require more complex solutions.
I realize that I don’t have very well-defined requirements – the purpose of my question here is to find out where these problems have been considered with more elegance.
What strategies exist for embedding change tracking in a structured data document format?
Thanks!
You might be interested in XML patch formats (e.g. as described by rfc 5261).
You could for example, build a list of such patches embedded at the top of your structured XML file and annotating each patch with its author, potential feature request/bug number and so on, potentially annotating it with semantic level patch information (such as “added such object”, “removed such rule”…). Using such a format could help you obtain old versions from your document rather easily as tools exist to treat it.