Pingback 1.0

This version:
http://www.hixie.ch/specs/pingback/pingback-1.0
Latest version:
http://www.hixie.ch/specs/pingback/pingback
Previous versions:
http://www.hixie.ch/specs/pingback/pingback-0.9.3
http://www.hixie.ch/specs/pingback/pingback-0.9.2
http://www.hixie.ch/specs/pingback/pingback-0.9.1
http://www.hixie.ch/specs/pingback/pingback-0.9
http://www.kryogenix.org/writings/tech/pingback
http://www.kryogenix.org/days/000138.cas
Editors:
Stuart Langridge <sil@kryogenix.org>
Ian Hickson <ian@hixie.ch>

Abstract

Pingback is a method for web authors to request notification when somebody links to one of their documents. Typically, web publishing software will automatically inform the relevant parties on behalf of the user, allowing for the possibility of automatically creating links to referring documents.

For example, Alice writes an interesting article on her Web log. Bob then reads this article and comments about it, linking back to Alice's original post. Using pingback, Bob's software can automatically notify Alice that her post has been linked to, and Alice's software can then include this information on her site.

Status of This Document

This is a stable specification. Comments are welcome on the blogite mailing list (archived).

There are currently six known separate implementations of this specification, although no formal testing has been done to check how compliant they are:

Authors are encouraged to pingback http://www.hixie.ch/specs/pingback/pingback to register their implementations.

Available languages

The English version of this specification is the only normative version. However, for translations of this document, see http://www.hixie.ch/specs/pingback/translations/. Currently available translations are:

Errata

This section was added after the final publication date of the specification.

(2007-01-16) In order to avoid susceptibility to denial of service attacks, pingback servers that fetch the specified source document (as described in section 3) are urged to impose limits on the size of the source document to be examined and the rate of data transfer. Thanks to Blake Matheny for bringing this issue to our attention.

Table of Contents


1. Introduction

The pingback system is a way for a blog to be automatically notified when other Web sites link to it. It is entirely transparent to the linking author, requiring no user intervention to work, and operates on principles of automatic discovery of everything that it needs to know. A sample blog post involving pingback might go like this:

  1. Alice posts to her blog. The post she's made includes a link to a post on Bob's blog.
  2. Alice's blogging system contacts Bob's blogging system and says "look, Alice made a post which linked to one of your posts".
  3. Bob's blogging system then includes a link back to Alice's post on his original post.
  4. Reader's of Bob's article can follow this link to Alice's post to read her opinion.

It enables reverse linking — a way of going back up a chain of links rather than merely drilling down.

1.1. Technical Details

The pingback mechanism uses an HTTP header and an HTML or XHTML <link> element for autodiscovery, and uses a single XML-RPC call for notifying the target site of the link on the source site.

It is intended that compliant pingback clients and pingback servers be implementable with minimal effort using libraries typically available in CGI environments. For this reason, the requirements on parsing HTTP headers and HTML documents have been kept to a strict minimum.

1.2. Definitions

source URI
The address of the entry on the site containing the link.
pingback client
The software that establishes the connection to inform the server about the link from the source to the target. Typically, the source will be the client.
pingback-enabled resource
A document, image or other resource that advertises a pingback server using a pingback HTTP header or a pingback link element.
pingback server
The software that accepts XML-RPC connections. Typically, the target URI will be associated with the server (e.g. on the same host).
pingback user agent
A single system, which is both a pingback client and a pingback server.
target URI
The target of the link on the source site. This SHOULD be a pingback-enabled page.

The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", "SHOULD", "SHOULD NOT", "RECOMMENDED", "MAY", and "OPTIONAL" in this document are to be interpreted as described in [RFC 2119].


2. Server Autodiscovery

There are two mechanisms for the automatic discovery of pingback servers: HTML (or XHTML) <link> elements and HTTP headers. A pingback-enabled resource MUST either be served with an X-Pingback HTTP header or contain a <link> element, or both. Pingback-enabled HTML and XHTML pages MUST be valid. Clients MAY refuse to search invalid pages for pingback information.

Note that how the client is told of the source and target URIs is out of the scope of this specification. Typically blogs will extract external links from posts being made to find the target URIs.

2.1. HTTP Header

Pingback-enabled resources MAY be returned with a X-Pingback HTTP header. For example, a PNG image served with the following headers would be pingback-enabled:

HTTP/1.1 200 OK
Date: Sun, 08 Sep 2002 15:05:37 GMT
Server: Apache/1.3.26 (Unix)
Last-Modified: Thu, 28 Dec 2000 03:18:26 GMT
ETag: "65044-15b9c-3a4ab102"
Accept-Ranges: bytes
Content-Length: 88988
Connection: close
Content-Type: image/png
X-Pingback: http://charlie.example.com/pingback/xmlrpc

.PNG...
  

The value of the X-Pingback header MUST be the absolute URI of the pingback XML-RPC server.

Pages MUST NOT include more than one such header. HTML and XHTML documents MAY include a <link> element in addition to an HTTP header, although this is discouraged. If included, the header SHOULD have exactly the same value as the <link> element. In the case of a discrepancy, the HTTP header SHALL override the <link> element, however, authors should be aware that some clients will not process HTTP headers due to limitations of their environment.

Pingback-enabled resources MUST NOT use the HTTP Link header for advertising pingback servers. HTTP Link headers require non-trivial parsing, and were therefore deemed too heavy-duty for the purposes of pingback server autodiscovery.

2.2. Link Element

An HTML or XHTML pingback-enabled page MAY contain a <link> element in one of the following two forms:

HTML
<link rel="pingback" href="pingback server">
XHTML
<link rel="pingback" href="pingback server" />

If used, the link element MUST match the appropriate form exactly (including the whitespace before the slash, for instance).

Pages MUST NOT include more than one such element, and MUST NOT include such a string matching the pattern described below unless it is intended to be the link element.

The pingback server placeholder MUST be replaced by the absolute URI of the pingback XML-RPC server. This URI MUST NOT include entities other than &amp;, &lt;, &gt;, and &quot;. Other characters that would not be valid in the HTML document or that cannot be represented in the document's character encoding MUST be escaped using the %xx mechanism as described in [RFC2396].

These strict requirements are intended to drastically reduce the requirements on clients implementing server autodiscovery, as it was deemed that requiring clients to implement an HTML parser in addition to an XML parser was a too heavy burden, given how easy it would be for page authors to comply to the restrictions described above.

2.3. Autodiscovery Algorithm

Pingback clients, given a source URI and a target URI, SHOULD fetch the target URI and follow the following steps to find the pingback server URI.

  1. Examine the HTTP headers of the response. If there are any X-Pingback headers then the first such header's value should be used as the pingback server URI. Clients MUST examine the HTTP headers if they are able to. If for some reason the HTTP headers are not available to the implementation then this step MAY be skipped, however, implementers should be aware that this will reduce the usefulness of their application as link elements cannot be used for resources that are neither HTML nor XHTML, and HTTP headers are defined to override link elements when they differ.
  2. Otherwise, search the entity body for the first match of the following regular expression:
    <link rel="pingback" href="([^"]+)" ?/?>
  3. If the regular expression matched, clients MUST then expand the four allowed entities (&amp; for &, &lt; for <, &gt; for >, and &quot; for ").

Having extracted this pingback server URI, it SHOULD be used to send an XML-RPC request as described below.

If the there is no X-Pingback header and the regular expression does not match, then the target in question does not support pingback as defined by this specification and the client MAY do whatever it likes. However, it is RECOMMENDED that clients do not attempt to be more lenient (e.g. by correctly parsing the HTML and looking for <link> elements that look like pingback links from an HTML point of view) because this will lead to some systems recognising the link and others ignoring it.

Clients MAY optimise the search. For example:

Note, however, that these optimisations are prone to being caught out by legitimate documents, for example those having comments containing the strings given above, or those with large inline stylesheets appearing before the pingback link. Authors are encouraged to take these possible optimisations into account when deciding where to place their pingback links.


3. XML-RPC Interface

Pingback clients, having discovered a pingback server, SHOULD send the server an XML-RPC request with the method name pingback.ping and two arguments, the source URI and the target URI respectively. [XML-RPC]

pingback.ping
Notifies the server that a link has been added to sourceURI, pointing to targetURI.
Parameters
sourceURI of type string
The absolute URI of the post on the source page containing the link to the target site.
targetURI of type string
The absolute URI of the target of the link, as given on the source page.
Return Value
A string, as described below.
Faults
If an error condition occurs, then the appropriate fault code from the following list should be used. Clients can quickly determine the kind of error from bits 5-8. 0×001x fault codes are used for problems with the source URI, 0×002x codes are for problems with the target URI, and 0×003x codes are used when the URIs are fine but the pingback cannot be acknowledged for some other reaon.
0
A generic fault code. Servers MAY use this error code instead of any of the others if they do not have a way of determining the correct fault code.
0×0010 (16)
The source URI does not exist.
0×0011 (17)
The source URI does not contain a link to the target URI, and so cannot be used as a source.
0×0020 (32)
The specified target URI does not exist. This MUST only be used when the target definitely does not exist, rather than when the target may exist but is not recognised. See the next error.
0×0021 (33)
The specified target URI cannot be used as a target. It either doesn't exist, or it is not a pingback-enabled resource. For example, on a blog, typically only permalinks are pingback-enabled, and trying to pingback the home page, or a set of posts, will fail with this error.
0×0030 (48)
The pingback has already been registered.
0×0031 (49)
Access denied.
0×0032 (50)
The server could not communicate with an upstream server, or received an error from an upstream server, and therefore could not complete the request. This is similar to HTTP's 402 Bad Gateway error. This error SHOULD be used by pingback proxies when propagating errors.
In addition, [FaultCodes] defines some standard fault codes that servers MAY use to report higher level errors.

Servers MUST respond to this function call either with a single string or with a fault code.

If the pingback request is successful, then the return value MUST be a single string, containing as much information as the server deems useful. This string is only expected to be used for debugging purposes.

If the result is unsuccessful, then the server MUST respond with an RPC fault value. The fault code should be either one of the codes listed above, or the generic fault code zero if the server cannot determine the correct fault code.

Clients MAY ignore the return value, whether the request was successful or not. It is RECOMMENDED that clients do not show the result of successful requests to the user.

Upon receiving a request, servers MAY do what they like. However, the following steps are RECOMMENDED:

  1. The server MAY attempt to fetch the source URI to verify that the source does indeed link to the target.
  2. The server MAY check its own data to ensure that the target exists and is a valid entry.
  3. The server MAY check that the pingback has not already been registered.
  4. The server MAY record the pingback.
  5. The server MAY regenerate the site's pages (if the pages are static).

4. Conformance Requirements

To claim conformance to this specification a pingback client MUST support server autodiscovery as described in this specification and MUST correctly send pingback XML-RPC calls.

To claim conformance to this specification a pingback server MUST be able to receive pingback XML-RPC calls and MUST always return results that conform to the allowed return values. Returning detailed (non-zero) fault codes is OPTIONAL.

Note that some pingback servers may not have associated pages. For example, a pingback gateway server could be standalone, and other pages would then use the link element to link to this gateway server instead of providing a server of their own. To claim conformance to this specification a pingback-enabled resource MUST have either an HTTP X-Pingback header or a link element in order to allow for server autodiscovery.

To claim conformance to this specification a pingback user agent MUST support server autodiscovery as described in this specification, MUST correctly send pingback XML-RPC calls, MUST be able to receive pingback XML-RPC calls, MUST always return results that conform to the allowed return values (returning detailed (non-zero) fault codes is OPTIONAL), and MUST have either an HTTP X-Pingback header or a link element on all potential target pages in order to allow for server autodiscovery.


5. Example

Here is a more detailed look at what could happen between Alice and Bob during the example described in the introduction.

  1. Alice posts to her blog. The post she's made includes a link to a post on Bob's blog. The permalink to Alice's new post is http://alice.example.org/#p123, and the URL of the link to Bob's blog is http://bob.example.net/#foo.
  2. Alice's blogging system parses all the external links out of Alice's post, and finds http://bob.example.net/#foo.
  3. It then requests the first 5 kilobytes of the page referred to by the link.
  4. It looks for an X-Pingback header, but fails to find one.
  5. It scans this page fragment for the pingback link tag, which it finds:
    <link rel="pingback" href="http://bob.example.net/xmlrpcserver">
    If this tag had not been contained in the page, then Bob's blog would not support pingback, so Alice's software would have given up here (moving on to the next link found in step 2).
  6. Next, since the link was there, it executes the the following XML-RPC call to http://bob.example.net/xmlrpcserver:
    pingback.ping('http://alice.example.org/#p123', 'http://bob.example.net/#foo')
  7. Alice's blogging system repeats step 3 to 6 for each external link that was found in the post.

There ends the work undertaken by Alice's system. The rest of the work is performed by Bob's blog.

  1. Bob's blog receives a ping from Alice's blog (the ping sent in step 6 above), naming http://alice.example.org/#p123 (the site linking to Bob) and http://bob.example.net/#foo (the page Alice linked to).
  2. Bob's blog confirms that http://bob.example.net/#foo is in fact a post on this blog.
  3. It then requests the content of http://alice.example.org/#p123 and checks the Content-Type of the entity returned to make sure it is text of some sort.
  4. It verifies that this content does indeed contain a link to http://bob.example.net/#foo (to prevent spamming of pingbacks).
  5. Bob's blog also retrieves other data required from the content of Alice's new post, such as the page title, an extract of the page content surrounding the link to Bob's post, any attributes indicating which language the page is in, and so forth.
  6. Finally, Bob's post records the pingback in its database, and regenerates the static pages referring to Bob's post so that they mention the pingback.

A. References

[RFC 2119]
Key words for use in RFCs to Indicate Requirement Levels, S. Bradner. IETF, March 1997. RFC 2119 is available at http://www.normos.org/ietf/rfc/rfc2119.txt.
[RFC 2396]
Uniform Resource Identifiers (URI): Generic Syntax, T. Berners-Lee, R. Fielding, L. Masinter. IETF, August 1998. RFC 2396 is available at http://www.normos.org/ietf/rfc/rfc2396.txt
[XML-RPC]
XML-RPC Specification, D. Winer. UserLand Software, Inc, June 1999. XML-RPC is available at http://www.xmlrpc.com/spec
[FaultCodes]
Specification for Fault Code Interoperability, D. Libby, et al. May 2001. The Specification for Fault Code Interoperability is available at http://xmlrpc-epi.sourceforge.net/specs/rfc.fault_codes.php