Why read-only Wikipedia mirrors are not enough

What Turkey's Wikipedia Block Showed About Copies, Corrections, and Censorship-Resistant Knowledge

Censorship-resistant knowledge is usually pictured as a storage problem: make enough copies and no government can erase them. The two and a half years when Wikipedia was blocked in Turkey suggest a harder lesson. Copies of the encyclopedia did escape the block, but readers inside the country lost something a copy cannot give back, which is the ability to join the conversation, argue with sources, and correct the articles that triggered the ban. This page serves learning and research, freedom technology, and the move from scarcity toward abundance of knowledge, by asking what a free encyclopedia needs beyond a mirror that can be read.

What happened in Turkey

On April 29, 2017, at 8 in the morning local time, every language edition of Wikipedia became unreachable for people in Turkey. According to the later press release of the Turkish Constitutional Court, a government security office asked the national telecommunications authority to remove two pieces of content, to block those pages if they were not removed, and to block the entire site if blocking single pages was not possible. The authority blocked the whole site because, in its view, blocking by single address was not technically possible. The two disputed texts described Turkey as a major outside actor in the Syrian civil war.

On December 26, 2019, the court held that the block violated freedom of expression. The Guardian reported a vote of 10 to 6. Access began to return on January 15, 2020, after the full ruling was published.

The copy that escaped

Within days, the team behind IPFS, a peer-to-peer file system that finds files by a fingerprint of their contents rather than by the server that holds them, published a full snapshot of Turkish Wikipedia. Anyone with the snapshot could serve it, readers could check that the files had not been altered, and the copy did not depend on any single domain name. The Turkish edition needed about 10 GB of storage, and the team estimated that English would need about 250 GB.

The same announcement was candid about what it had built. It set two goals. The first was periodic read-only snapshots. The second, a full read and write Wikipedia running on the peer-to-peer network, was described as much more difficult and dependent on close work with the Wikimedia Foundation. Only the first goal was reached.

What a copy cannot do

The most striking line in this story comes from the court itself. After the block began, volunteer editors elsewhere made extensive changes to the disputed texts, reworded them more carefully, and removed claims and sources they found unreliable. The court then noted that the block had also kept users in Turkey from discussing the disputed content and had denied active editors the chance to adjust, change, and contribute to it.

In other words, the articles at the center of the dispute were revised during the very period when the people with the most direct stake in them were shut out. A snapshot freezes an encyclopedia at one moment. It cannot carry the talk pages, the challenges to a weak source, or the slow process by which a contested article improves. A reader holding a mirror holds a record of the past argument, not a seat in the present one.

This matters for anyone who thinks of knowledge as a growing commons. Abundance of facts depends on abundance of correction. A country cut off from editing does not only lose access. The rest of the world loses its local experts, its native speakers, and its witnesses, and the shared record becomes poorer for everyone.

Why the mirror went quiet

The second lesson is about maintenance. The IPFS snapshot repository is now archived, and its own documentation explains why. Unpacking a Kiwix archive into tens of millions of small files took days, needed a dedicated machine, and broke whenever the archive format changed. Successive snapshots shared almost no data, so keeping a history meant paying full price each time. The README adds that in practice one person with a big machine ran the build, which is why the snapshots stopped.

The archived project now recommends a different design: keep the compressed archive file that Kiwix already publishes intact, place it on the peer-to-peer network as a single file, and read it in a browser that fetches only the parts it needs. The lesson is general. A censorship-resistant system that depends on one dedicated volunteer is only as resistant as that volunteer's free time. Durable tools ride on formats and pipelines that many people already use.

The quieter infrastructure that lasts

Two less dramatic projects explain why Wikipedia stays hard to erase. Kiwix, a nonprofit based in Lausanne, packages Wikipedia and other educational sites into compressed files that can be read entirely offline, and it states that more than half of the world's population lacks reliable internet access because of cost, infrastructure, or censorship. The Wikimedia Foundation publishes free database dumps of its wikis, used by researchers, offline readers, and archives, and it asks volunteers to host mirrors so that as many copies as possible exist.

These dumps are what make every other mirror possible. The IPFS team built its Turkish snapshot from a Kiwix archive. Openly licensed text plus routine public exports is the real foundation. Clever distribution sits on top of it.

A three-layer view of censorship-resistant knowledge

The Turkish case suggests that a knowledge commons needs three layers, and that tools tend to solve them in order of ease.

• Reading. Offline archives, peer-to-peer snapshots, and physical copies already handle this well when the content is openly licensed.
• Verifying. Content fingerprints let a reader confirm that a copy matches what was published, which matters when copies pass through many hands.
• Participating. Proposing edits, challenging sources, and reaching agreement across borders remains the weakest layer. It needs identity, moderation, and a way to merge changes made in separate places.

The third layer is where future work on decentralized wikis, permanent storage, and censorship-resistant social identity has the most to offer, and also where it is least mature.

Honest limits

This page relies on a court press release and a foundation statement rather than the full Turkish judgment, and on project documentation for the technical history. It does not measure how many people in Turkey used the peer-to-peer copy, Kiwix, or other workarounds during the block, because no reliable figure was found. Decentralized editing raises real problems of vandalism, spam, and coordinated manipulation that a single trusted host handles more easily. Nothing here is legal advice or guidance on evading the law in any particular country.

Discussion questions, essay ideas, and learning related AI prompt ideas

• Explain why the Turkish Constitutional Court treated the loss of editing as a separate harm from the loss of reading.
• Compare the costs of keeping a read-only mirror current with the costs of letting people in a blocked country propose edits safely.
• Describe how a decentralized wiki could merge edits made while two groups of editors could not reach each other.
• Analyze why a censorship-resistant project that depends on one volunteer can fail even when its technology works.
• Evaluate whether open licenses and public data exports do more for knowledge abundance than any single distribution tool.
• Design a study that would estimate how many readers in a blocked country actually use offline or peer-to-peer copies.

See also

• Sneakernets against censorship
• Permanent decentralized wiki design
• Arweave storage pricing
• Nostr social identity
• Offline AI tutors
• Backing up civilization