Subject: The Trove project -- next-generation Internet software archiving Date: Wed, 17 Jun 1998 13:21:13 GMT The Trove project is aimed at producing the next-generation Internet software archive facility. Its objective is to replace the ad-hoc hacks now used at Sunsite, CPAN, PSA, the Red Hat contrib directory and elsewhere with a flexible and powerful Web-accessible database that will, among other things, support unified searches across all Trove archives. As the person who does most of the package-filing work on Sunsite, I became convinced some months ago that the classical FTP-tree model of Internet software archive just doesn't cut it any more. It's way too intensive of maintainer time, making it unsustainable as Linux explodes in popularity. It also makes software excessively hard to find unless you're a regular archive user with a map of the site in your head. The thin WWW wrappers in place at Sunsite and CPAN make those archives a little prettier, but don't solve the underlying scaling problems. The Python guys over at PSA do a bit better on the user side through effective leveraging of free-text searching, but haven't solved the maintainer-intensiveness problem. And none of the ad-hoc archives talk to or can search each other. We need a fundamentally new approach, and we need it before these vital pieces of community infrastructure collapse under their own weight. I consider a solution to this problem critical enough that Trove is now my top priority after continuing the public push for the open-source development mode. The Trove project has been up and running for month or so. The design has firmed up nicely and working code is beginning to appear. If you think you might be interested in helping solve these problems, please look over the Trove project page at: http://www.tuxedo.org/~esr/trove There's a design document accessible in HTML there that describes the project's objectives and present status in detail. I'm aiming high. I want the maintainers of the major existing archives to buy in early, so that by year-end the present creaking infrastructure can be replaced with something better. I can deliver Sunsite myself. The Python community's leaders are already on board, and so is Red Hat. And we've had an expression of interest from CPAN's maintainer. Anybody else running an archive site is urged to join now so we can be sure to accomodate your community's needs in the design. We could use more design input (especially from people running archives now), and we'll need a fair bit of implementation and testing help. CGI and Web experts would be particularly welcome, also people who grok PGP and web-session-authentication problems. -- Eric S. Raymond, http://www.tuxedo.org/~esr