This document describes the protocol used by the Google Data APIs, including information about what a query looks like, what results look like, and so on.

For more information about the Google Data APIs, see the Google Data Developer's Guide document and the Protocol Guide.

Audience

This document is intended for anyone wanting to understand the details of the XML format and protocol used by the Google Data APIs.

If you just want to write code that uses the Google Data client APIs, then you don't need to know these details; instead, you can use the language-specific client libraries.

But if you want to understand the protocol, read this document. For example, you may want to read this document to help you with any of the following tasks:

• evaluating the Google Data architecture
• coding using the protocol without using the provided Google Data client libraries
• writing a client library in a new language

This document assumes that you understand the basics of XML, namespaces, syndicated feeds, and the GET, POST, PUT, and DELETE requests in HTTP, as well as HTTP's concept of a "resource." For more information about those things, see the Additional resources section of this document.

This document doesn't rely on any particular programming language; you can send and receive Google Data messages using any programming language that lets you issue HTTP requests and parse XML-based responses.

Protocol details

This section describes the Google Data document format and query syntax.

Document format

Google Data, Atom, and RSS 2.0 all share the same basic data model: a container that holds both some global data and any number of entries. For each protocol, the format is defined by a base schema, but it can be extended using foreign namespaces.

The Google Data APIs can use either the Atom syndication format (for both reads and writes) or the RSS format (for reads only).

Atom is Google Data's default format. To request a response in RSS format, use the /alt=rss/ parameter; for more information, see Query requests.

When you request data in RSS format, Google Data supplies a feed (or other representation of the resource) in RSS format. If there's no equivalent RSS property for a given Google Data property, Google Data uses the Atom property, labeling it with an appropriate namespace to indicate that it's an extension to RSS.

Note: Most Google Data feeds in Atom format use the Atom namespace as the default namespace by specifying an xmlns attribute on the feed element; see the examples section for examples of how to do that. Thus, the examples in this document don't explicitly specify atom: for elements in an Atom-format feed.

The following tables show the Atom and RSS representations of the elements of the schema. All data not mentioned in these tables is treated as plain XML and shows up the same in both representations. Unless indicated otherwise, the XML elements in a given column are in the namespace corresponding to that column. This summary uses standard XPath notation: in particular, slashes show the element hierarchy, and an @ sign indicates an attribute of an element.

In each of the following tables, the highlighted items are required.

The following table shows the elements of a Google Data feed:

The following table shows the elements of a Google Data search-results feed. Note that Google Data exposes some of the OpenSearch 1.1 Response elements in its search-results feeds.

The following table shows the elements of a Google Data entry:

Queries

This section describes how to use the query system.

Query model design tenets

The query model is intentionally very simple. The basic tenets are:

• Queries are expressed as HTTP URIs, rather than as HTTP headers or as part of the payload. One benefit of this approach is that you can link to a query.
• Predicates are scoped to a single item. Thus, there's no way to send a correlation query such as "find all emails from people who sent me at least 10 emails today."
• The set of properties that queries can predicate on is very limited; most queries are simply full text search queries.
• Result ordering is up to the implementation.
• The protocol is naturally extensible. If you want to expose additional predicates or sorting in your service, you can do so easily through the introduction of new parameters.

Query requests

A client queries a Google Data service by issuing an HTTP GET request. The query URI consists of the resource's URI (called FeedURI in Atom) followed by query parameters. Most query parameters are represented as traditional ?name=value[&...] URL parameters. Category parameters are handled differently; see below.

For example, if the FeedURI is http://www.example.com/feeds/jo, then you might send a query with the following URI:

http://www.example.com/feeds/jo?q=Darcy&updated-min=2005-04-19T15:30:00Z

Google Data services support HTTP Conditional GET. They set the Last-Modified response header based upon the value of the <atom:updated> element in the returned feed or entry. A client can send this value back as the value of the If-Modified-Since request header to avoid retrieving the content again if it hasn't changed. If the content hasn't changed since the If-Modified-Since time, then the Google Data service returns a 304 (Not Modified) HTTP response.

A Google Data service must support category queries and alt queries; support for other parameters is optional. Passing a standard parameter not understood by a given service results in a 403 Forbidden response. Passing an unsupported nonstandard parameter results in a 400 Bad Request response. For information on other status codes, see the HTTP status codes section of this document.

The standard query parameters are summarized in the following table. All parameter values need to be URL encoded.

About category queries

We decided to specify a slightly unusual format for category queries. Instead of a query like the following:

http://example.com/jo?category=Fritz&category=2006

we use:

http://example.com/jo/-/Fritz/2006

This approach identifies a resource without using query parameters, and it produces cleaner URIs. We chose this approach for categories because we think that category queries will be the most common queries.

The drawback to this approach is that we require you to use /-/ in category queries, so that Google Data services can distinguish category queries from other resource URIs, such as http://example.com/jo/MyPost/comments.

Query responses

Queries return an Atom feed, an Atom entry, or an RSS feed, depending on the request parameters.

Query results contain the following OpenSearch elements directly under the <feed> element or the <channel> element (depending on whether results are Atom or RSS):

openSearch:totalResults

The total number of search results for the query (not necessarily all present in the results feed).

openSearch:startIndex

The 1-based index of the first result.

openSearch:itemsPerPage

The maximum number of items that appear on one page. This allows clients to generate direct links to any set of subsequent pages. However, for a possible pitfall in using this number, see the note regarding start-index in the table in the Query requests section.

The Atom response feed and entries may also include any of the following Atom and Google Data elements (as well as others listed in the Atom specification):

<link rel="http://schemas.google.com/g/2005#feed" type="application/atom+xml" href="..."/>

Specifies the URI where the complete Atom feed can be retrieved.

<link rel="http://schemas.google.com/g/2005#post" type="application/atom+xml" href="..."/>

Specifies the Atom feed's PostURI (where new entries can be posted).

<link rel="self" type="..." href="..."/>

Contains the URI of this resource. The value of the type attribute depends on the requested format. If no data changes in the interim, sending another GET to this URI returns the same response.

<link rel="previous" type="application/atom+xml" href="..."/>

Specifies the URI of the previous chunk of this query result set, if it is chunked.

<link rel="next" type="application/atom+xml" href="..."/>

Specifies the URI of the next chunk of this query result set, if it is chunked.

<link rel="edit" type="application/atom+xml" href="..."/>

Specifies the Atom entry's EditURI (where you send an updated entry).

Here's a sample response body, in response to a search query:

<?xml version="1.0" encoding="UTF-8"?>
<feed xmlns:atom="http://www.w3.org/2005/Atom"
        xmlns:openSearch="http://a9.com/-/spec/opensearchrss/1.0/">
  <id>http://www.example.com/feed/1234.1/posts/full</id> 
  <updated>2005-09-16T00:42:06Z</updated> 
  <title type="text">Books and Romance with Jo and Liz</title> 
  <link rel="alternate" type="text/html" href="http://www.example.net/"/> 
  <link rel="http://schemas.google.com/g/2005#feed"
    type="application/atom+xml"
    href="http://www.example.com/feed/1234.1/posts/full"/> 
  <link rel="http://schemas.google.com/g/2005#post"
    type="application/atom+xml"
    href="http://www.example.com/feed/1234.1/posts/full"/> 
  <link rel="self" type="application/atom+xml"
    href="http://www.example.com/feed/1234.1/posts/full"/> 
  <author>
    <name>Elizabeth Bennet</name> 
    <email>liz@gmail.com</email> 
  </author>
  <generator version="1.0"
    uri="http://www.example.com">Example Generator Engine</generator> 
  <openSearch:totalResults>2</openSearch:totalResults> 
  <openSearch:startIndex>0</openSearch:startIndex> 
  <entry>
    <id>http://www.example.com/feed/1234.1/posts/full/4521614025009481151</id> 
    <published>2005-01-09T08:00:00Z</published> 
    <updated>2005-01-09T08:00:00Z</updated> 
    <category scheme="http://www.example.com/type" term="blog.post"/> 
    <title type="text">This is the title of entry 1009</title> 
    <content type="xhtml">
      <div
        xmlns="http://www.w3.org/1999/xhtml">This is the entry body of entry 1009</div> 
    </content>
    <link rel="alternate" type="text/html"
      href="http://www.example.com/posturl"/> 
    <link rel="edit" type="application/atom+xml"
      href="http://www.example.com/feed/1234.1/posts/full/4521614025009481151"/> 
    <author>
      <name>Elizabeth Bennet</name> 
      <email>liz@gmail.com</email> 
    </author>
  </entry>
  <entry>
    <id>http://www.example.com/feed/1234.1/posts/full/3067545004648931569</id> 
    <published>2005-01-07T08:00:00Z</published> 
    <updated>2005-01-07T08:02:00Z</updated> 
    <category scheme="http://www.example.com/type" term="blog.post"/> 
    <title type="text">This is the title of entry 1007</title> 
    <content type="xhtml">
      <div
        xmlns="http://www.w3.org/1999/xhtml">This is the entry body of entry 1007</div> 
    </content>
    <link rel="alternate" type="text/html"
      href="http://www.example.com/posturl"/> 
    <link rel="edit" type="application/atom+xml"
      href="http://www.example.com/feed/1234.1/posts/full/3067545004648931569"/> 
    <author>
      <name>Elizabeth Bennet</name> 
      <email>liz@gmail.com</email> 
    </author>
  </entry>
</feed>

If the requested feed is in the Atom format, if no query parameters are specified, and if the result doesn't contain all the entries, the following element is inserted into the top-level feed: <link rel="next" type="application/atom+xml" href="..."/>. It points to a feed containing the next set of entries. Subsequent sets contain a corresponding <link rel="previous" type="application/atom+xml" href="..."/> element. By following all the next links, a client can retrieve all entries from a feed.

HTTP status codes

The following table describes what various HTTP status codes mean in the context of the Google Data services.

Optimistic concurrency (versioning)

Sometimes it is important to ensure that multiple clients don't inadvertently overwrite one another's changes. This is most easily accomplished by ensuring that the current version of an entry that a client is modifying is the same as the version that the client is basing its modifications on. If a second client makes an update before the first client does, then the first client's update is denied, because the first client is no longer basing its modifications on the latest version.

In Google Data feeds that support versioning, we achieve these semantics by appending a version ID to each entry's EditURI. Note that only the EditURI is affected, not the entry ID. In this scheme, each update changes the entry's EditURI, thus guaranteeing that subsequent updates based on the original version fail. Deletes, of course, are identical to updates with respect to this feature; if you send a delete with an old version number, the delete fails.

Not all Google Data feeds support optimistic concurrency. In a feed that does support it, if the server detects a version conflict on PUT or DELETE, the server responds with 409 Conflict. The body of the response contains the current state of the entry (an Atom entry document). The client is advised to resolve the conflict and resubmit the request, using the EditURI from the 409 response.

Motivation and design notes

This approach to optimistic concurrency allows us to implement the semantics we want without requiring new markup for version IDs, which makes Google Data's responses compatible with non-Google Data Atom endpoints.

Instead of specifying version IDs, we could have chosen to look at the update timestamp on each entry (/atom:entry/atom:updated). However, there are two problems with using the update timestamp:

• It only works for updates and not deletions.
• It forces applications to use date/time values as version IDs, which would make it harder to retrofit Google Data on top of many existing data stores.

Authentication

When a client tries to access a service, it may need to provide the user's credentials to the service, to demonstrate that the user has the authority to perform the action in question.

The approach that a client should use for authentication depends on the type of client:

• A desktop application should use a Google-specific authentication system called Account Authentication for Installed Applications (also known as "ClientLogin"). (Web-based clients should not use this system.)
• A web-based client, such as a third-party front end to a Google Data service, should use a Google-specific authentication system called Account Authentication Proxy for Web-Based Applications (also known as "AuthSub").

In the ClientLogin system, the desktop client asks the user for their credentials, and then sends those credentials to the Google authentication system.

If authentication succeeds, then the authentication system returns a token that the client subsequently uses (in an HTTP Authorization header) when it sends Google Data requests.

If authentication fails, then the server returns a 403 Forbidden status code, along with a WWW-Authenticate header containing a challenge applicable to the authentication.

The AuthSub system works similarly, except that instead of asking the user for their credentials, it connects the user to a Google service that requests credentials. The service then returns a token that the web application can use; the advantage of this approach is that Google (rather than the web front end) securely handles and stores the user's credentials.

For details about these authentication systems, see the Google Data Authentication Overview or the Google Account Authentication documentation.

Session state

Many business logic implementations require session stickiness—keeping track of the state of a user's session.

Google tracks session state in two ways: using cookies, and using a token that can be sent as a query parameter. Both methods achieve the same effect. We recommend that clients support one of these session-state tracking methods (either one is sufficient). If a client doesn't support either of these methods, then that client will still work with Google Data services, but performance may suffer compared to clients that do support these methods. Specifically, if a client doesn't support these methods, then every request results in a redirect, and therefore every request (and any associated data) is sent to the server twice, which affects the performance of both the client and the server.

The Google client libraries handle session state for you, so if you use our libraries, you don't have to do anything to get session state support.

Additional resources

You may find the following third-party documents useful:

• Comparison of Atom and RSS from intertwingly
• Overview of Atom from IBM
• Dublin Core extensions to RSS
• HTTP 1.1 method definitions; specification for GET, POST, PUT, and DELETE
• HTTP 1.1 status code definitions
• How to Create a REST Protocol
• Building Web Services the REST Way
• A Technical Introduction to XML
• XML Namespaces by Example

Back to top