Araby.com

» see araby.com as it lived, on the wayback machine

~ the goodbye ~

Araby.com never reached Turkey at any scale and there is no reason it should have. it was language-coded rather than region-coded, built for readers of Arabic, and i am not one. what makes it worth a grave from where i sit is that Turkey ran the same experiment in the same years with Arabul and Netbul, and lost it the same way.

it launched in September 2006 out of Maktoob Group, which by then was the largest Arabic site on the web, a portal and webmail service out of Jordan with millions of accounts on it. Araby was their search engine, and the claim they made for it was that it was the first one built for Arabic from the inside rather than an English engine with Arabic bolted onto the front.

that claim had teeth. Arabic is heavily inflected, words carry prefixes and suffixes that change the surface form of a root, and short vowels usually are not written at all. a search engine that treats the string literally will miss most of what a reader would consider the same word. handling that properly is a real engineering problem and in 2006 the large engines had not put much effort into it. Araby had a few hundred thousand people using it within months.

~ the chronicle ~

~ the language problem ~

this is the part i recognize from a completely different alphabet. Turkish is agglutinative, which means a single root can carry a stack of suffixes and turn into a word an English-shaped index will never match against its own root. the pre-Google Turkish engines of exactly this period were written up in the local information-retrieval papers for their stemming work, because stemming was the whole game.

every one of those regional engines made the same bet. the bet was that language handling was hard enough to be a moat, that a company which understood the grammar would beat a foreign one that did not. it was a reasonable bet in 2004. it was wrong by about 2008, because the foreign one had the crawl and the ranking already, and could fix its own stemming at leisure. language turned out to be the cheapest of those problems to solve.

~ how it actually ended ~

in August 2009 Yahoo bought Maktoob, which was the big regional exit of the decade and the reason anybody outside the Middle East knows the name. the deal was for the portal and the audience. it explicitly did not include Araby, and it did not include Souq or cashU either. those went into a new holding company called Jabbar Internet Group.

being left out of a deal is a diagnosis. Souq survived being left out because it was a business with revenue, and Amazon eventually bought it. Araby had no revenue anybody wanted and no chance left against Google in Arabic. the archive shows a real site through June 2010 and an address that stopped serving its own homepage the month after. no announcement, as usual.

~ what got buried with it ~

the specific technical work is what got lost. somebody spent years on Arabic morphological analysis for search, and the output of that went into a portal sale and then into a holding company that had no use for it. the same thing happened in Turkish and in half a dozen other languages with awkward grammar. a shelf of dead engines, each one holding real linguistic engineering that never got reused.

Maktoob itself did not last much longer as a name. Yahoo ran it as Yahoo Maktoob, shut the Dubai office in 2016 and finally stopped publishing under the brand at the start of 2023. the portal outlived its search engine by thirteen years and both of them are gone now, which is a fairly ordinary ending for the regional web of that decade.

@}-->--

              )
             (_)
            .-'-.
            |   |
            |   |
            |   |
            |   |
          __|   |__   .-.
       .-'  |   |  `-:   :
      :     `---'     :-'
       `-._       _.-'
           '""""""

« previous  ·  random  ·  next »