The Tethys Sea is gone. What replaced it — the Mediterranean in miniature, and a desert — is the record of its disappearance. But remnants persist: in the shape of *Dracaena cinnabari* on a limestone plateau, in the grammatical architecture of a language spoken by 70,000 people, in the soil moisture beneath a forest that does not know it is the last of its kind. The Tethys is gone, but its shadows still exist, and they are now being contested by forces moving faster than geology.
Chapter 1 described the conditions of preservation. Chapter 2 must describe the active work of erasure.
···
**The Infrastructure Arrives**
The Yemen civil war that began in 2015 brought Socotra into a geopolitical contest it had largely avoided. The island was occupied by UAE forces in May 2018 — a deployment that Yemeni officials called an invasion and that the UAE described as a response to a request for humanitarian assistance following a cyclone. Within two years, the UAE had built new roads, expanded the airport, and stationed troops and equipment in numbers that transformed the island's physical landscape more rapidly than any prior development.
There are arguments for this infrastructure. The roads opened the interior. The airport expansion enabled larger flights. The humanitarian case was not entirely fabricated — Cyclone Mekunu had caused real damage.
But infrastructure is never neutral. Roads that open the interior also open the interior. An expanded airport that enables larger aid flights also enables larger commercial flights, more tourists, more arrivals who don't speak Socotri. Infrastructure determines which connections become possible, and possibility has a way of becoming habit, and habit has a way of becoming identity.
The UAE brought something else with it: a media ecosystem. Arabic-language satellite television, Arabic-language content on mobile phones, and — most consequentially — Etisalat, the UAE telecom provider, whose SIM cards began to circulate on an island where connectivity had previously been intermittent and limited.
In January 2026, the Southern Transitional Council, which the UAE had backed as a political presence on the island, officially dissolved. The UAE military completed its withdrawal. The formal apparatus of occupation — the flag, the troops, the political structures — packed up and left.
Etisalat did not.
Tour guides still recommend, as of late 2025, purchasing an Etisalat SIM at Abu Dhabi Airport before flying to Socotra. The signal is better. The service is more reliable. The digital infrastructure outlasted the military infrastructure, because digital infrastructure is cheaper to maintain, invisible to withdrawal negotiations, and vastly more effective at the things that matter for cultural continuation.
Use ← and → to turn chapters
·
·
·
**The Telecom Ghost**
Think about what it means for a language to survive the internet.
Socotri has survived Arabic contact for centuries — it survived medieval Arab traders on the coastal routes, it survived the Ottoman period, it survived the British protectorate in Aden that nominally included Socotra. It survived because contact was limited, because the island's interior remained difficult to reach, because the language of daily life and work and love and argument remained Socotri in the absence of a compelling reason to change it.
The internet eliminates the compelling-reason problem. It doesn't require you to travel to the mainland to access mainland culture. It delivers it to your pocket, at low latency, continuously, for the rest of your life, beginning in adolescence.
The question of which language Socotri teenagers speak to each other in private is not determined by flag. It is not determined by which government formally administers the island. It is determined by which apps load fast, which content is cheap to stream, which memes circulate in which language, which keyboard their phone defaults to, which language autocomplete has learned from years of text message history.
Etisalat's continued presence on Socotra means continued embedding in the UAE telecom ecosystem: UAE app stores, UAE content delivery networks, UAE default language settings. The military left. The content recommendation algorithm remained. The military's departure will show up in political science papers. The content recommendation algorithm's effect will show up in the language patterns of children who are currently five years old.
This is the telecom ghost: the digital control apparatus that outlasts the physical one, that operates below the threshold of withdrawal negotiations, that doesn't require soldiers or flags or formal authority to do its work.
···
**50 Downloads**
There is a Socotri keyboard. It was built — a careful piece of engineering, covering the phonological peculiarities of the language, the emphatic consonants, the pharyngeal sounds, the vowel contrasts that Arabic orthography cannot easily represent. It was released.
It has approximately 50 downloads.
This is, on the surface, a story about a keyboard. Beneath the surface, it is a story about what linguists sometimes call the chicken-and-egg problem of minority language digitization. There is no content to type with the keyboard. Without content to type, there is no reason to install the keyboard. Without the keyboard installed, there is no efficient way to produce content. The content ecosystem precedes the tool, and the tool precedes the content ecosystem, and neither can bootstrap the other by existing in isolation.
What is striking about the 50-download figure is what it is *not* evidence of. It is not evidence that Socotri speakers don't love their language. The Socotri Poetry Festival, documented by Stanford University Press, took place in 2011 and drew pastoralists from across the island to compete in oral verse performance — in Socotri, publicly, with the full seriousness of a tradition that knew its own worth. The community that produced that festival is not a community that needs to be taught to value its language. It is a community that has loved its language for as long as it can remember, because the language is the medium in which its memory exists.
The keyboard problem is not a passion problem. It is a structural problem. What is missing is not desire. What is missing is the scaffolding that turns desire into sustainable practice: content to read, platforms to post on, communities that reward participation, institutional recognition that makes using the language feel like a forward-looking act rather than a backward-looking one. The keyboard is a tool for a workshop that hasn't been built yet.
Compare this to what Etisalat provides: a complete content ecosystem, a social network, an app store, a billion pieces of Arabic-language content delivered at high speed to a device in a teenager's hand. The asymmetry is not between a good technology and a bad technology. It is between an engineering solution operating in a vacuum and an engineering solution backed by the full weight of a state-sponsored telecommunications monopoly.
···
**The Naming War**
In September 2024, UNESCO hosted the first unified-alphabet workshop in Hadibo. Among the participants was the Yemeni Ambassador to Djibouti, who noted in his remarks that the workshop represented a "qualitative shift" in classifying Soqotri as one of the "Southern Arabic languages."
Every specialist in Modern South Arabian Languages — the family to which Socotri belongs — says, with varying degrees of patience, that there are no longer any grounds for this classification. Socotri is not Arabic. It does not descend from Arabic. It predates Arabic contact in the region by a chronological margin so large as to make the comparison almost incoherent. The Modern South Arabian languages as a group — Socotri, Mehri, Shehri, Harsusi, Bathari, Jibbali — are classified as a distinct branch of the Semitic family tree, separated from Arabic at the earliest stages of Semitic divergence.
But UNESCO reproduced the Ambassador's framing not in a footnote, not with a cautionary gloss, but in body text. It was printed as context. It was archived as record.
This is how naming wars are won, and this is why they matter. The classification of a language as a dialect of a dominant language is not merely a taxonomic error. It is a consequential political move that weakens the case for the language's independent preservation. If Socotri is a "Southern Arabic language," then its survival is already ensured: Arabic is not endangered. Its unique features — the grammatical fossils that preserve the structure of proto-Semitic in ways Arabic cannot — become dialectal quirks rather than irreplaceable windows into the history of a language family. Its speakers become members of a large, thriving speech community rather than the last custodians of something singular.
Biologists understand this dynamic. When a population is reclassified as a subspecies rather than a full species, it drops in conservation priority. The case for allocating resources to its preservation weakens. The urgency shifts. The Socotri naming dispute is operating in exactly this space — a space where the difference between "language" and "dialect," between "independent branch" and "Southern Arabic variety," has direct consequences for whether anyone funds documentation, whether any UNESCO mechanism recognizes it as distinctly endangered, whether anyone outside the island's 70,000 speakers has a institutional reason to care.
Tadewos Mekonnen, a linguist studying the broader dynamics of minority language endangerment, has noted that institutional misclassification is one of the most underappreciated extinction mechanisms. The species doesn't die because someone decided to kill it. It dies because someone filed it incorrectly, and the filing persisted, and the corrective work was never done.
The Yemeni Ambassador was not wrong to advocate for his framing. Ambassadors are advocates. The question is why UNESCO's record-keeping allowed advocacy to settle as fact — and what corrective work has been done since. The answer, as of this writing, is not encouraging.
···
**Soft Power with a Corpus**
In 2021, the Russian linguist Vitaly Naumkin published a major corpus of Socotri language materials. The corpus was funded, in part, by the UAE Embassy in Moscow.
By 2024, HSE University in Moscow had established a Centre for South Arabian Studies — a trilateral Russia-UAE academic partnership focusing on the languages and cultures of southern Arabia, including Socotri.
This is worth sitting with.
Academic documentation of endangered languages is among the most genuinely valuable work that linguistics produces. Naumkin's work on Socotri is real scholarship by real scholars, and the language materials it has produced are irreplaceable. None of this is in dispute.
What is also true is that language documentation is soft power, and soft power knows this about itself. Who documents a language determines how it is framed. Who archives it determines who controls the archive. Who publishes the corpus determines the citation network, the academic lineage, the institutional ownership of the record. When the UAE Embassy in Moscow funds Socotri documentation and the resulting corpus is published with Russian-UAE institutional affiliations, the scholarly record of the language becomes part of a specific geopolitical investment.
The engineering serves the politics. This doesn't make the engineering bad. It makes the politics worth examining, and it raises a question that the documentation community has not fully answered: what does it mean for the *community* of Socotri speakers — the 70,000 people on the island whose language is being documented — that the primary academic record of their language is being maintained by institutions whose interest in Socotra is substantially geopolitical?
The Soqotri Poetry Festival pastoralists didn't need Russian or UAE mediation to produce oral verse of extraordinary complexity. The community's relationship to its own language predates and exceeds every external intervention. What it currently lacks is the structural conditions — institutions, platforms, resources — that would allow that relationship to sustain itself in the presence of a telecom ghost and a misclassification in a UNESCO record.
···
**Hydrological Damage**
There is something uniquely precise about how the dragon blood trees are being damaged that deserves its own paragraph.
Each mature *Dracaena cinnabari* injects water into the surrounding soil at several times the rate of local rainfall. The umbrella canopy collects fog and channels it down the trunk, concentrating moisture that would otherwise evaporate before reaching the ground. On the Firmihin plateau, the dragon blood woodland is not merely a scenic attraction — it is the primary mechanism by which the water table is maintained. It is, in technical terms, hydrological infrastructure.
UAE extraction of timber and resin, and the disruption of the highland plateau that accompanied military and commercial activity, did not just damage trees. It damaged water. In a landscape where water is the scarce variable, where rainfall is measured in the low hundreds of millimeters per year, the removal of the fog-harvesting infrastructure is a form of dehydration. The damage doesn't announce itself as catastrophe. It accumulates quietly: a slightly lower water table, slightly drier soil, slightly less favorable conditions for seedling establishment, slightly higher juvenile mortality.
The military left. The hydrological damage did not. The trees that were removed will not be replaced in any human lifetime — *Dracaena cinnabari* grows slowly, and seedlings are devastated by the introduced goat and donkey populations whose numbers expanded during the period of disruption. The water table that was lowered will recover, if it recovers, over decades. The compounding effects will be felt long after the soldiers, the flags, the political structures, and the telecom SIM cards have been replaced by something else.
This is the other ghost: not the digital one, but the hydrological one. The fingerprint of an occupation written not in political records but in soil moisture data, in the age-class distribution of a tree population, in the gap between current precipitation and current water table.
···
**The Window**
Starlink is now operational on Socotra. A tour operator, 2Socotra.com, offers satellite internet at its camps — the first connectivity on the island that is genuinely independent of the UAE telecom infrastructure. You can now, on Socotra, connect to the internet without going through an Etisalat tower.
This is small. It is also important.
The combination of political transition — UAE military withdrawal, STC dissolution, the attendant instability and renegotiation — and the arrival of satellite-independent internet creates a window that did not exist before. A window for what? For documentation. For content creation in Socotri. For the Socotri keyboard to find its first hundred users who are not linguists. For platforms that serve Socotri-language content to be established before the content ecosystem calcifies around Arabic. For the institutional record-keeping to catch up to the reality that this is a distinct language with a distinct community that has a distinct and genuine interest in its own continuation.
The window is time-limited. Political transitions are short. Starlink access is currently expensive and accessed primarily by foreign tourists, not by the island's residents. The telecom ghost is entrenched and has years of installed infrastructure and user behavior behind it. The next generation of Socotri children is growing up now, forming their language habits now, deciding implicitly and continuously which language they live in.
Twenty million years of isolation produced a tree that bleeds red and a language that contains grammatical features preserved from before the Arabian Peninsula was a peninsula. The Tethys Sea closed. The forest fragmented. The survivors found themselves at the edges of a distribution that no longer had a middle.
What the mountain dialect speakers of Socotra's interior now represent — according to fieldwork reports from researchers who have visited — is a final layer of preservation within the preservation. The mountain Socotri that is least exposed to Arabic influence, least affected by the coastal fishing economy's integration with the mainland, is diverging from the coastal dialect in ways that will be irrecoverable within fifteen to twenty years. After that, it will be available only in recordings, if recordings have been made.
Twenty million years of preservation. Fifteen to twenty years of remaining window for the mountain dialect.
The Tethys Sea is gone. The forest that grew along its shores is gone. What remain are two trees — one in Socotra, one in the Canary Islands — connected by the memory of a forest neither of them knew. And a language, spoken in the mountains of an island in the Arabian Sea, whose grammatical structure is a fossil record of a proto-Semitic that no one speaks any more and no one will ever speak again.
The telecom ghost is patient. The Starlink window is not.
The question is whether anyone acts on it before the window closes — before the content ecosystem calcifies, before the mountain dialect becomes a recording, before the last speaker of the full register of Socotri takes that register with her into the ground like a Miocene seed that no longer has the conditions it needs to germinate.
The Telecom Ghost · Two Survivors from the Tethys Sea · Latent Press