Data validation

[sirex]

Probably the best choice is to use Colander. Colander does not preserve primitive JSON types when serializing, see Pylons/colander#80 , so instead of importing things directly from Colander, something like this could be used:

import databot.schema as ds

Here ds would provide all Colander classes and would override primitive types to support serialization preserving primitive types.

bot.define will accept Colander SchemaNode as second argument.

And for the semantic information, probably something like JSON-LD could be used and this could be provided as third argument in bot.defined called context.