Netscape Netcaster

~ the eulogy ~

Push assumed the line was already open. That is the whole story of this thing and of the entire year it belonged to. In 1997 an industry convinced itself that the future was content arriving on its own schedule rather than yours, and built the software for it before the connection that would make it sane existed anywhere except a corporate LAN.

Netcaster was Netscape's entry. It was announced under the codename Constellation, appeared in the Communicator 4 line from June 1997 and reached most Windows users in the 4.02 release, and it did two things. It let you subscribe to channels, meaning sites that agreed to be delivered rather than visited, and it gave you a webtop, a full-screen layer that sat over your desktop and showed you the channel whether you had asked for it right then or not.

My first modem was a 33.6 in 1998, on a Turkish line where the phone company charged by the minute and the connection was a decision you made and then ended. Software that wanted to sit connected and pull down updates on a timer was not a bad product for me, it was an expensive one. Half the world was in that position and the entire push category was designed as though nobody was.

~ the dossier ~

~ how it worked ~

A channel was ordinary web content with a schedule attached. You picked from a list, ten by default and expandable to twenty, and set how often each one should refresh. Netcaster fetched in the background and cached, so the content was on your disk when you looked at it. Your channel selections lived on a server rather than in the machine, which meant your setup followed you to a different computer, an idea that took the rest of the industry roughly fifteen years to arrive at again.

Underneath it Netscape had licensed Marimba's Castanet, the Java delivery system built by people who had come out of the original Java team. Castanet worked on a transmitter and tuner model where the tuner ran locally and pulled on a schedule, so what looked to the user like a broadcast being pushed at them was a poll they had configured once and forgotten about. Almost everything sold as push in 1997 was doing this.

~ the collapse ~

The competition was Microsoft's Active Channels in Internet Explorer 4 and PointCast, and all three were fighting over a market that turned out not to be there. PointCast was already notorious in corporate IT for saturating office networks, which is the part everyone remembers. The part that mattered more was that home users on metered dial-up had no reason to subscribe to anything, and the sites building channels had to author a second version of themselves in somebody's channel format to reach an audience nobody could measure.

Netscape had bigger problems by then anyway. Communicator 4.5 shipped in October 1998 without Netcaster in it, in the same period the company was losing the browser war outright and being bought by AOL. The feature did not get a farewell. It was cut from a release, the way features are.

~ what it became ~

Everything it wanted came true about six years later and under a different name. Subscribing to a site that then arrives without you visiting it is RSS, and RSS won by being a file format instead of a client, so nobody had to install Netscape's version of the idea. Settings that follow you between machines became the browser account. Background fetch and offline cache became the service worker.

So push was not wrong. It was early by exactly the length of the broadband rollout, which is the same explanation that covers most of the failures of 1997. The webtop is the one piece nobody has ever wanted since. Every few years a company tries again to own the whole screen instead of a window on it, and every time the answer comes back the same: the desktop belongs to the person sitting at it.

@}-->--

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

« previous  ·  random  ·  next »