IBM Home Page Reader

~ the obituary ~

The documented language list runs english, french, german, italian, japanese and spanish, with more added later. Turkish was never on it, and I am sighted, so there were two separate reasons this program was never going to sit on anything I owned. I found out what it was years after it came off sale, from reading about how screen readers actually work, which is something most people who build web pages learn late or never.

It launched in Japan in 1997 and reached Windows 95 in english the following year. The work behind it was Chieko Asakawa's at the IBM Tokyo research lab. She had injured an optic nerve at eleven in a swimming accident and was completely blind by fourteen, and she presented the interface design at the ACM accessibility conference in 1998. The name is literal. It read the home page to you.

The competition at the time worked by reading the screen. A screen reader would take whatever the operating system had drawn, interpret the pixels and the widget tree, and speak the result. That works acceptably for a word processor with one window and one column of text. It falls apart on a web page, where content scrolls out of view, sits inside frames, and carries meaning in structure that never shows up as anything visible.

Home Page Reader skipped the screen. It took the HTML itself and built its model of the document from that, which meant it knew what was a heading and what was a link before anything was painted. IBM kept it standards-clean on purpose, meeting the W3C's HTML 4.01 spec along with the first Web Content Accessibility Guidelines. It ran on top of Netscape Navigator 3.01 or better, on a 166 MHz machine with 32 megabytes of memory, over a 28.8k modem.

~ the case file ~

~ how it worked ~

The daily experience was the numeric keypad. Nearly everything a user did went through it, and the payoff for reading the markup was that a keystroke could jump between form fields or step through the tables and frames of a page instead of walking the document line by line from the top. There was a fast-forward key for skimming. Anybody who has listened to a screen reader grind through the same menu bar for the fortieth time on one site understands why that mattered.

The other half of the design was voice, and this is the part people who used it still describe first. Links were read in a female voice and plain text in a male one, with the user free to reassign which voice handled which kind of element. It carried in sound the thing a sighted reader gets from a color and an underline, without stopping to say the word link before every one of them.

~ the inheritance ~

In 2006 somebody from IBM posted to the Home Page Reader mailing list that there were no plans for further updates. Steve Faulkner wrote it up in november under the title IBM Home Page Reader is Dead, and the product came off sale that december. Version 3.04, from 2005, was the last one.

What killed it is that its own argument won. By the mid 2000s the browsers exposed accessibility interfaces of their own, the mainstream screen readers had learned to work from document structure rather than from pixels, and the guidelines IBM had helped write in public were the thing everyone tested against. A separate paid browser that spoke HTML made sense in 1998 because nothing else did it. It stopped making sense once everything did.

Asakawa stayed at IBM and was named an IBM Fellow in 2009. The habit her browser started, that assistive software should read the document rather than the drawing of the document, is where every accessibility tool now begins, and it is the reason the alt attribute you skipped filling in has consequences.

@}-->--

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

« previous  ·  random  ·  next »