This is "track", a system that tries to maintain software and data on multiple machines. Updates are made on a central machine (or machines) called the librarians, and made available for export. Receiving machines, called subscribers, initiate the transfer of files, by asking for the list of changed files and getting the files they need/want. Librarians have complete control over the files they want to export, the subscribers they want to export to and the times they are willing to make changes known, subscribers have complete control over the files they want to import, where they want to put those files and the times they want to import files. That allows track to be used between administratively independent domains, and makes it considerably more flexible and efficient than rdist. This software was originally written at Bell Communications Research by Daniel Nachbar. It's been modified a lot at the University of Toronto. We use it heavily on Suns running SunOS4.x and 5.3, Ultrix machines (Vaxen and DS3100s) and SGI Iris4Ds running Iris [45].x. In particular, it's enabled us to avoid YP, er, Network Information Service. UofT changes that I know of: - Inline find instead of invoking /bin/find for every line in a subscription list. - Made it work with all fields on the track line. - Fixed a few bugs, including a couple of nasty security holes. - Made trackd lint. John Linderman at AT&T provided some lint modes and general fixes for calltrack, and trackd. - Some performance improvements, and porting to different platforms. - SunOS5 port by Edwin Allum. We're not responsible for this. But we'd like to hear of improvements, bugs and fixes. - moraes@cs.toronto.edu (Mark Moraes, University of Toronto) - and hacks@cs.toronto.edu