A proposed simple specification for syncing tabular data over HTTP via JSON transaction logs.
(Originally drafted explicitly by Christopher Groskopf of the Panda Project and Francis Irving of ScraperWiki on 21st December 2011. Concept by Max Ogden. Named by Isaac Z. Schlueter)
Motivation document by Max Ogden: https://gist.github.com/1246804
Can be any URL, and is a GET request.
There are some extra URL parameters that a client can include:
- since - seq id to start at, returns everything after that, defaults to the start
- limit - maximum number of changes to return, defaults to unlimited XXX can a server optionally limit this to a maximum?
- include_data - if present and false, the wire protocol’s data field isn’t included
JSON representing part of the transaction log of changes to the tabular data.:
- seq - the sequence in the transaction log
- id - unique identifier of the row
- deleted - if present and true, means the row is being deleted by the transaction
- data - new data for the row, should not be present if deleted is true (also see include_data in the HTTP endpoint section)
Keys in the data dictionary can be any string, with any UTF-8 character, or whitespace in them. To meet JSON specificaiton, they must be quoted if not simple like a variable name.
Note that as it is JSON, the keys can be in any order and change order.
Values in the data dictionary must be one of these types:
- Strings, in quotes
- Numbers, integers or floats
- Dates, a string containing an ISO 8601 date or date/time. Only in UTC no timezones.
Optional - by convention name variables storing WGS 84 latitude/longitude so they end _lat and _lng. e.g.
Return error messages in a JSON dictionary with the key error, value an arbitary string.:
Should we just use http://syncable.org/ ?
Make it use the right MUST type words from RFCs to be clear what is compulsory.