Search help

The archive search box accepts plain words, quoted phrases, boolean operators, prefix matches, and full Message-IDs. Here's how each one behaves with examples you can click through.

Basic terms

Type one or more words and every result will contain all of them (they're AND-joined). Hyphens and apostrophes are treated as word characters. Everything else is dropped as punctuation.

You type	Interpreted as	Example
`namespaces`	`namespaces`	try it
`xml namespaces`	`xml AND namespaces`	try it
`relax-ng`	`relax-ng` (hyphen preserved)	try it

Quoted phrases

Put words in double quotes to match them as an exact phrase.

You type	Interpreted as	Example
`"xml namespace"`	phrase match	try it
`"content negotiation" REST`	phrase AND `rest`	try it

Boolean operators

Uppercase AND, OR, and NOT are passed through to the full-text search engine. Use them sparingly — most queries don't need them.

You type	Interpreted as	Example
`xpath OR xquery`	either term	try it
`xslt NOT xslt2`	mentions of XSLT without the string "xslt2"	try it

Prefix / wildcard

Append * to a word to match anything starting with it.

You type	Matches	Example
`namesp*`	namespaces, namespace, namespacing…	try it
`schemat*`	schematron, schematic, schemata…	try it

Message-ID lookup

If your input looks like a Message-ID (something@host, optionally <…>-wrapped), the archive bypasses full-text search and jumps straight to that specific message if it exists.

You type	Behaviour
`<[email protected]>`	direct lookup, 301 to message page
`[email protected]`	same, angle brackets optional

Filters

Filters are appended as query-string parameters. You can combine any of them with the main search term.

Parameter	Example	Meaning
`year`	`?q=namespaces&year=2003`	only 2003
`month`	`?q=namespaces&month=200301`	only Jan 2003
`from`	`?q=namespaces&from=Tim Bray`	author name or email contains
`from_date`	`?q=…&from_date=2003-01`	lower bound (YYYY, YYYY-MM, YYYY-MM-DD, YYYYMM)
`to_date`	`?q=…&to_date=2003-06-30`	upper bound
`sort`	`?q=…&sort=relevance`	`newest` (default), `oldest`, or `relevance`
`list`	`?q=…&list=xml-dev`	restrict to one mailing list (multi-list sites)
`page`	`?q=…&page=3`	pagination
`limit`	`?q=…&limit=500`	results per page (max 1000, default 200)

Walkthrough with screenshots

The four screenshots below show typical research workflows against this archive. Each one is taken live from the same URL you see beneath it.

Home page of the xml-dev archive showing stats, sparkline, and top contributors — **1. Landing page.** Activity sparkline across 23 years, top contributors, biggest threads, and the month table below. Click any bar in the sparkline to jump directly to that month.

Search results for namespaces showing message list with date, author, subject, and snippet — **2. Keyword search.** `?q=namespaces` — default sort is newest first. Each row shows date, author, subject, and a snippet with highlighted matches.

Search results filtered by year=2003 and from=Tim Bray — **3. Structured filters.** `?q=namespaces&year=2003&from=Tim Bray` — only messages by Tim Bray posted in 2003 that contain the word "namespaces".

Thread view for one of the biggest 2002 discussions on xml-dev — **4. Thread view.** All messages in a single conversation, grouped by canonical subject regardless of `[xml-dev]` or `Re:` prefix variations. Prev/next jump to adjacent messages in the thread, then spill over into the next thread in the month.

Query rewrites and warnings

When the search box changes what you typed in a way you might care about (punctuation stripped, slashes turned into phrases, etc.) a warning appears above the results explaining the change. The most common warnings are:

"Punctuation and special characters were ignored"
"Slash-separated terms were treated as phrases"
"Your query had no searchable terms after punctuation was stripped"

This is the archive's way of being honest about what it actually searched for. Historical mailing lists are full of strings like xmlns:foo="http://…" that don't survive tokenisation; the warning tells you when that's happening.

API

Every search URL has a JSON counterpart at /api/archive/search with the same parameters. Example:

curl 'https://xml.org/api/archive/search?q=namespaces&year=2003&limit=5' | jq

The response includes total_count, offset, sort, warnings, and the list of matching messages with snippets. Designed for programmatic research and LLM tooling.