Pavatar (2006–2011): the Personal Avatar protocol

The idea

In 2006 many blogs showed commenter pictures from one central service, Gravatar, and the project's own home page described that server as “not working”. Pavatar proposed the opposite: every commenter hosts their own avatar on their own website, and the blog software discovers it from the website address the commenter already typed into the comment form. No sign-up, no central server, and no company collecting a profile of where you comment.

The rules were deliberately small. A Pavatar is an 80×80 pixel GIF, PNG or JPEG, no larger than 409,600 bits, served with the correct Content-Type.

How autodiscovery works

A website could publish its Pavatar in three ways, checked in this order of precedence:

HTTP header — X-Pavatar: http://example.com/path/my-pavatar.png
Link element — <link rel="pavatar" href="http://example.com/path/my-pavatar.png" /> in the page head
Direct URL — a file named pavatar.png next to the page or at the site root, “the favicon.ico way”

The keyword none in the header or link meant “I do not want an avatar shown”. Implementations were expected to cache avatars and use conditional requests to keep traffic low. Pavatar explained walks through the algorithm with a diagram, and the Pavatar check page gives the same steps as commands you can run today.

The specification

The specification was written and edited by Jeena Paradies (© 2006). The first drafts, 0.1.0–0.1.2, appeared on jeenaparadies.net; version 0.2.0 moved to pavatar.com in May 2006, and version 0.3.0 was published as a Candidate Recommendation on 18 December 2006 — “believed to be stable” and open for implementation. Each version is explained on its own page: Pavatar explained, 0.3.0 and 0.2.0.

Implementations

By 2008 the project home page listed implementations in several languages:

a “Check your Pavatar!” script written in Ruby,
the Block weblog software (C),
a WordPress plugin (PHP) and a DotClear module (PHP),
a “universal plugin” that worked like Gravatar but on your own server (Ruby),
the Perl module Net::Pavatar on CPAN.

Step-by-step guides explained how to publish a Pavatar from hosted blogs that did not let you upload files freely: on Blogger you added the <link rel="pavatar"> tag right after <head> in the blog template, and a similar guide covered Xanga. The Blogger guide was written by Kurt McKee and released under the GNU Free Documentation License. Links to the implementations you can still find are on Pavatar explained.

The author and the community

The protocol was created and maintained by Jeena Paradies — the author of the specification, the main voice on the support forum and the contact for the spec at the time. The project also had a support forum from early 2007, where developers debated naming (rel="avatar" versus rel="pavatar"), Gravatar compatibility and OpenID integration.

What happened to pavatar.com

Before May 2006
First drafts of the specification, versions 0.1.0–0.1.2, on jeenaparadies.net.
May 2006
Version 0.2.0 and a project site on pavatar.com, with a “Check your Pavatar!” script (announcement, in German).
18 December 2006
Version 0.3.0, a Candidate Recommendation open for implementation.
2007–2012
The support forum, plugins for blog software, discussions on the microformats mailing list and about OpenID integration. The original home page is still in the Wayback Machine in January 2012.
2013–2023
The Wayback Machine shows a short project page in 2015 that pointed to a GitHub project, and otherwise error responses: HTTP 403 in 2013 and 2016–2018, server errors in 2021–2023. The domain eventually expired.
15 January 2024
David S bought the expired domain at an auction. David is not connected to Jeena Paradies or the original Pavatar project.
2024–2025
A generic tech blog by the same owner, unrelated to Pavatar. It has been removed, and its addresses return “410 Gone”.
Since 2026
This archive and a new site about data tools and automations. Pavatar-era addresses (/spec/, /check/, /forum/…) lead to the matching archive pages.

Who runs this site now

I'm David S. I bought this domain in January 2024, ran a general tech blog here in 2024–25 and took it down; this is the second, focused version. I keep this archive so that links from 2006–2011 lead to an honest record of the protocol instead of an error page. I had no part in Pavatar itself, and nothing here speaks for its author.

The rest of the site is about data tools and automations — guides, n8n workflows and free tools. More about me is on the About page. If you were part of the Pavatar community and something here is wrong or missing, please get in touch.