Hi all,
I wanted to flag an update from the Wikimedia Data Engineering team in case it’s relevant to your work, especially for those of you who may rely on Wikimedia XML dumps for research.
In short:
- A new set of MediaWiki Content File Exports is now available, providing unparsed content from Wikimedia’s public wikis in XML format. - There are two monthly datasets: - mediawiki_content_history: https://dumps.wikimedia.org/other/mediawiki_content_history/ - Full revision history for all pages - mediawiki_content_current: https://dumps.wikimedia.org/other/mediawiki_content_current/ - Latest revision only for each page - This change was made because the legacy dump infrastructure at https://dumps.wikimedia.org/backup-index.html has struggled to reliably generate XML exports for larger wikis. - The older XML dump pipeline is now considered deprecated, though SQL dumps will continue and some legacy generation may persist temporarily.
You can read the full announcement here: https://lists.wikimedia.org/hyperkitty/list/xmldatadumps-l@lists.wikimedia.o..., and view the full documentation at: https://wikitech.wikimedia.org/wiki/MediaWiki_Content_File_Exports.
Best, Kinneret
Thanks for the update.
Il 09/02/26 13:56, Kinneret Gordon via Wiki-research-l ha scritto:
You can read the full announcement here: https://lists.wikimedia.org/hyperkitty/list/xmldatadumps- l@lists.wikimedia.org/thread/E6D5EU4PMSTSOI2J7A46HJ3YW2W554CS/, and view the full documentation at: https://wikitech.wikimedia.org/wiki/MediaWiki_Content_File_Exports.
How was the filename structure decided? Was there a reason to drop the previous conventions? For example, the two files itwiki-2026-02-01-p2p4267764.xml.bz2 and itwiki-2026-02-01-p100066p103578.xml.bz2 are named very similarly, but one contains full history and the other only current versions.
Best, Federico
Hi Federico,
I don't have insights into how that was handled. I've cc'd Xabriel Collazo Mojica from the Data Engineering team who can help answer this question. You can also follow up at https://phabricator.wikimedia.org/T414389 with your question.
Best, Kinneret
On Thu, Feb 12, 2026 at 12:08 PM Federico Leva (Nemo) nemowiki@gmail.com wrote:
Thanks for the update.
Il 09/02/26 13:56, Kinneret Gordon via Wiki-research-l ha scritto:
You can read the full announcement here: https://lists.wikimedia.org/hyperkitty/list/xmldatadumps- l@lists.wikimedia.org/thread/E6D5EU4PMSTSOI2J7A46HJ3YW2W554CS/, and view the full documentation at: https://wikitech.wikimedia.org/wiki/MediaWiki_Content_File_Exports.
How was the filename structure decided? Was there a reason to drop the previous conventions? For example, the two files itwiki-2026-02-01-p2p4267764.xml.bz2 and itwiki-2026-02-01-p100066p103578.xml.bz2 are named very similarly, but one contains full history and the other only current versions.
Best, Federico
Federico, I'm adding Xabriel's response below. Hope this is helpful.
The new file export treats a history export and a current revision export as two separate jobs, with two distinct URLs. Let's look at the fully qualified names of the two files that I think you refer to:
https://dumps.wikimedia.org/other/mediawiki_content_current/itwiki/2026-02-0... and https://dumps.wikimedia.org/other/mediawiki_content_history/itwiki/2026-02-0...
Thus under https://dumps.wikimedia.org/other/mediawiki_content_current/itwiki/2026-02-0..., you will find all the files associated with the current revision export.
Under https://dumps.wikimedia.org/other/mediawiki_content_history/itwiki/2026-02-0..., you will find all the files associated with the full history file export.
This is indeed different from the legacy dumps where all the files were dumped into the same folder, but it allows us to manage the jobs independently.
In a couple of places the 'all' in 'all revisions' is italicised. Does this mean that this dump includes oversighted revisions?
cheers stuart
-- ...let us be heard from red core to black sky
wiki-research-l@lists.wikimedia.org