[Draft] Clarify serialization and discoverability

content/docs/specifications/data-package.md

@@ -81,15 +81,19 @@ Several example data packages can be found in the [datasets organization on gith
 
 ### Descriptor
 
+On logical level, Data Package descriptor is represented by a data structure. The data structure `MUST` be a JSON-serializable `object` as defined in [RFC 4627](http://www.ietf.org/rfc/rfc4627.txt).
+
+On physical level, Data Package descriptor is represented by a file. A data producer `MAY` use any suitable serialization format and `SHOULD` name the file `datapackage.json`. A data consumer `MUST` support JSON serialization format and `MAY` support other serialization formats like YAML or TOML.
+
+The above states that JSON is the only serialization format that `MUST` be used for publishing a Data Package while other serialization formats can be used in projects or systems internally if supported by corresponding implementations.
+
+This specification does not define any discoverability mechanisms making a serialized Data Package be referenced only directly by its URI. It means that techically the name of a Data Package file is irrelevant although it is good practice to use `datapackage.json` as a public convention.
+
 The descriptor is the central file in a Data Package. It provides:
 
 - General metadata such as the package's title, license, publisher etc
 - A list of the data "resources" that make up the package including their location on disk or online and other relevant information (including, possibly, schema information about these data resources in a structured form)
 
-A Data Package descriptor `MUST` be a valid JSON `object`. (JSON is defined in [RFC 4627][]). When available as a file it `MUST` be named `datapackage.json` and it `MUST` be placed in the top-level directory (relative to any other resources provided as part of the data package).
-
-[RFC 4627]: http://www.ietf.org/rfc/rfc4627.txt
-
 The descriptor `MUST` contain a `resources` property describing the data resources.
 
 All other properties are considered `metadata` properties. The descriptor `MAY` contain any number of other `metadata` properties. The following sections provides a description of required and optional metadata properties for a Data Package descriptor.

content/docs/specifications/data-resource.md

@@ -81,9 +81,15 @@ A comprehensive Data Resource example with all required, recommended and optiona
 }
 ```
 
-### Descriptor
+## Descriptor
 
-A Data Resource descriptor `MUST` be a valid JSON `object`. (JSON is defined in [RFC 4627][]).
+On logical level, Data Resource descriptor is represented by a data structure. The data structure `MUST` be a JSON-serializable `object` as defined in [RFC 4627](http://www.ietf.org/rfc/rfc4627.txt).
+
+On physical level, Data Resource descriptor is represented by a file. A data producer `MAY` use any suitable serialization format. A data consumer `MUST` support JSON serialization format and `MAY` support other serialization formats like YAML or TOML.
+
+The above states that JSON is the only serialization format that `MUST` be used for publishing a Data Resource while other serialization formats can be used in projects or systems internally if supported by corresponding implementations.
+
+This specification does not define any discoverability mechanisms making a serialized Data Resource be referenced only directly by its URI.
 
 Key properties of the descriptor are described below. A descriptor `MAY` include any number of properties in additional to those described below as required and optional properties.
 

content/docs/specifications/table-dialect.md

@@ -37,7 +37,15 @@ CSV Dialect is useful for programmes which might have to deal with multiple dial
 
 Some related work can be found in [this comparison of csv dialect support](https://docs.google.com/spreadsheet/ccc?key=0AmU3V2vcPKrIdEhoU1NQSWtoQmJwcUNCelJtdkx2bFE&usp=sharing), this [example of similar JSON format](http://panda.readthedocs.org/en/latest/api.html#data-uploads), and in Python's [PEP 305](http://www.python.org/dev/peps/pep-0305/).
 
-## Specification
+## Descriptor
+
+On logical level, Table Dialect descriptor is represented by a data structure. The data structure `MUST` be a JSON-serializable `object` as defined in [RFC 4627](http://www.ietf.org/rfc/rfc4627.txt).
+
+On physical level, Table Dialect descriptor is represented by a file. A data producer `MAY` use any suitable serialization format. A data consumer `MUST` support JSON serialization format and `MAY` support other serialization formats like YAML or TOML.
+
+The above states that JSON is the only serialization format that `MUST` be used for publishing a Table Dialect while other serialization formats can be used in projects or systems internally if supported by corresponding implementations.
+
+This specification does not define any discoverability mechanisms making a serialized Table Dialect be referenced only directly by its URI.
 
 A CSV Dialect descriptor, `dialect`, `MUST` be a JSON `object` with the following properties:
 

content/docs/specifications/table-schema.md

@@ -67,7 +67,17 @@ For example, `constraints` `SHOULD` be tested on the logical representation of d
 
 ## Descriptor
 
-A Table Schema is represented by a descriptor. The descriptor `MUST` be a JSON `object` (JSON is defined in [RFC 4627](http://www.ietf.org/rfc/rfc4627.txt)).
+On logical level, Table Schema descriptor is represented by a data structure. The data structure `MUST` be a JSON-serializable `object` as defined in [RFC 4627](http://www.ietf.org/rfc/rfc4627.txt).
+
+On physical level, Table Schema descriptor is represented by a file. A data producer `MAY` use any suitable serialization format. A data consumer `MUST` support JSON serialization format and `MAY` support other serialization formats like YAML or TOML.
+
+The above states that JSON is the only serialization format that `MUST` be used for publishing a Table Schema while other serialization formats can be used in projects or systems internally if supported by corresponding implementations.
+
+This specification does not define any discoverability mechanisms making a serialized Table Schema be referenced only directly by its URI.
+
+## Metadata
+
+### Fields
 
 It `MUST` contain a property `fields`. `fields` `MUST` be an array where each entry in the array is a field descriptor (as defined below). The order of elements in `fields` array `SHOULD` be the order of fields in the CSV file. The number of elements in `fields` array `SHOULD` be the same as the number of fields in the CSV file.