Currently we have a couple of problems. Testing a) Testing is taking longer and longer. b) we do not test all variant combinations. We don't test release and we don't test different variants for runtime-link and other combinations that users mght use. c) Testers are making larger commitments. Many users download and install on variations of platforms we test on examples since last release of 1.32 a) versions of stlport other than 4.5.3 b) Mac OSX in release mode c) Metrowerks for MAC ? Basically, our process doesn't scale well. Each one of raises a couple of new issues. I would like to see the "Getting Started" section of the documentation somewhat re-organized in the following way: a) test the test subsystem - this results in a build of the test subsystem b) test each library - this results in a build by default c) generate compiler_status and link html file for the target environment c) stage- move libraries to final directory. That is, installing BOOST would by default include testing it. Any problems would be addressed by uploading test results to a known destination. I envision that the "test results" would be the compiler_status and links.html files. (Note I recently uploaded a version of compiler_status to the vault which builds tables of ALL the combinations tested) Hopefully all of the above could be accomplished from withing bjam rather then platform dependent shell or msdos scripts. This would have the following effects. a)Testing would automatically be performed on platforms that users actually use. b)Testing could be more exhaustive as its spread among a larger number of machines c)More combinations of variants could be tested. d)It would permit users with difficulties to be offered better help. Robert Ramey