Collected on 2026-07-31
Measurements
The rarity in this catalogue has always been the curators' opinion, and the site has said so since day one. These two columns are not: they come from outside, from sources anyone can check.
What is measured, and how much of it
- Interoperable since
- The date the last of the three engines started shipping the element, from MDN's browser-compat-data. Covers 30 of 30. Where the datum is an upper bound — BCD's
≤— the page says “no later than”, and there are 9 such cases. - Measured usage
- The fraction of visits in which Chrome sees the element on the page, from the Chrome Platform Status counters. Covers 8 of 30 — only elements with an exactly-named counter get in, with no guessing. It is one browser, and the sample is its own.
Where the measurement contradicts the curators
Comparing the 8 elements with measured usage pairwise, 26 of the 28 pairs confirm the order the curators had guessed. These do not:
I did not correct the rarity. The measurement is of usage in Chrome today, and the curators' label means something else — how much the element is remembered by people writing HTML. They are different rulers, and swapping one for the other without saying so would be worse than an opinion openly held. What could be done was to show both side by side.
The whole table
In order of arrival: from what was already ready in 2008 to what shipped yesterday.
| Specimen | Interoperable since | Measured usage | Rarity (opinion) |
|---|---|---|---|
<address> |
11/12/2008 | — | forgotten |
<map> |
11/12/2008 | 0.47% of visits | forgotten |
<noscript> |
11/12/2008 | — | forgotten |
<dl> |
no later than 08/06/2009 | — | common |
<abbr> |
08/06/2009 | — | uncommon |
<kbd> |
no later than 08/06/2009 | — | uncommon |
<samp> |
no later than 08/06/2009 | — | forgotten |
<small> |
no later than 08/06/2009 | — | common |
<var> |
no later than 08/06/2009 | — | forgotten |
<wbr> |
08/06/2009 | — | forgotten |
<del> |
no later than 08/06/2009 | — | uncommon |
<ins> |
no later than 08/06/2009 | — | uncommon |
<fieldset> |
no later than 08/06/2009 | — | common |
<optgroup> |
no later than 08/06/2009 | — | uncommon |
<hgroup> |
22/03/2011 | — | forgotten |
<figure> |
20/07/2011 | — | common |
<mark> |
20/07/2011 | — | uncommon |
<bdi> |
25/07/2012 | — | forgotten |
<progress> |
25/07/2012 | 1.86% of visits | uncommon |
<meter> |
09/10/2012 | 0.20% of visits | forgotten |
<output> |
22/10/2013 | — | forgotten |
<template> |
16/10/2014 | — | uncommon |
<ruby> |
12/05/2015 | — | forgotten |
<picture> |
21/03/2016 | — | common |
<details> |
20/09/2016 | 4.37% of visits | common |
<data> |
17/10/2017 | 0.48% of visits | forgotten |
<time> |
17/10/2017 | 8.16% of visits | common |
<datalist> |
25/03/2019 | 2.02% of visits | uncommon |
<dialog> |
14/03/2022 | 9.73% of visits | uncommon |
<search> |
10/10/2023 | — | forgotten |
None of this is fetched while the site is being generated. What talks to the network is coleta.py, which keeps the raw responses in dados/ — versioned, alongside the code. The build reads from the database and never leaves the machine, so two generations of the same commit give the same site, with or without internet.