WAIS

~ the eulogy ~

Searching used to start with a question that has since been deleted from the world, which is where. Not what, where. You picked a machine and connected to it, and whatever that machine held was the entire universe your query could reach. Ask the wrong server a perfect question and it politely told you it had nothing.

WAIS was built in 1990 at Thinking Machines in Cambridge, Massachusetts, jointly funded with Apple and Dow Jones alongside the accountancy firm KPMG Peat Marwick, and it ran on the company's Connection Machine supercomputers. The protocol borrowed from the library standard Z39.50 and then went its own way over TCP/IP, on port 210. Clients existed for Windows and Macintosh, for NeXT and X, for GNU Emacs, and for people with nothing but a text terminal.

The searching itself was ahead of where it had any right to be. You typed words rather than boolean operators, and results came back sorted by a relevance score rather than by date or by filename. Then came the part that still feels modern: you could point at a document you liked and ask the system for more like that one, and it would use the document itself as the query. Commercial search engines took most of a decade to offer that again.

The customers were serious. The Environmental Protection Agency ran a server and so did the Library of Congress, and later so did the Wall Street Journal and Encyclopaedia Britannica. Ross Perot's 1992 presidential campaign used it to connect field offices to the national office, which makes WAIS an early corporate intranet several years before that word existed.

I arrived after this argument had been settled. My first modem was in 1998, and by then a search meant one page with one box on it, and nobody had to know that servers were even a thing that could be chosen between. Reading about WAIS now is reading about a version of the internet whose basic assumption was that you already knew roughly where the answer lived. That assumption was true when the network was mostly institutions, and it stopped being true almost exactly when I got there.

~ the dossier ~

~ the directory of servers ~

Thinking Machines ran one server that solved the where problem by making it into another search. It was called the Directory of Servers, and it was an ordinary WAIS database whose records described other WAIS databases. A server operator uploaded a record about themselves, and anyone could then search the directory to find machines likely to hold what they were after.

Search for the thing that searches: this is the idea every internet directory since has been a version of, and it was already running before the web existed. The same design went straight into the Government Information Locator Service, which the United States wrote into federal law in the Paperwork Reduction Act of 1995. Public WAIS also ended up bolted onto Gopher as its full-text engine, because Gopher's own search only looked at menu titles.

~ where the people went ~

Thinking Machines filed for Chapter 11 on 17 August 1994, killed by a collapse in government supercomputer funding rather than by anything WAIS did. Brewster Kahle and Harry Morris had already left to spin WAIS Inc out into Menlo Park with Bruce Gilliat, and in May 1995 that company was sold to AOL for fifteen million dollars. Development scattered immediately after: one engineer's variant became MetaStar, another's eventually became ScienceServer and was sold to Elsevier.

The free branch outlived the commercial one on paper. The National Science Foundation funded CNIDR to build freeWAIS after Thinking Machines dropped support, and a group at the University of Dortmund extended that into freeWAIS-sf. None of it mattered by then, because Mosaic had arrived and the answer to where had become nowhere in particular.

Nobody switched WAIS off, so there is no date to put on this page with any honesty, only 1995 and the reasons around it. The best-known thing about it now is what its authors did next: Kahle and Gilliat left to found Alexa Internet and the Internet Archive. The man who built search for an internet too small to need a crawler spent the rest of his career keeping a copy of the one that replaced it.

@}-->--

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

« previous  ·  random  ·  next »