Robert Jones wrote:
I am unclear of what the vault contains, and of its relation to the sandbox.
The sandbox is a versioned repository, whereas the the vault appears not to be.
What is the workflow here, if there is one? By what path do libraries arrive at Boost proper (ie. part of some release), and where in the process are the review points?
Both the vault and the sandbox provide alternative means developers can use to make code still in development available: the vault provides zip file downloads while the sandbox provides full version control when that's more applicable. Developers producing code for Boost are free to use either, both or neither depending on preference :-) However, once a library has matured to the point that the author would like it to become an official part of Boost, then they ask for a formal review of the code (http://www.boost.org/development/submissions.html), and generally also make a zip file available for download from the vault (just because it's more convenient for potential reviewers that way). Once code has been accepted into Boost and is in mainline SVN Trunk, some of us continue to use the sandbox for general development as well. To pick an example the Boost.Math special functions/distribution code has a branch in the sandbox: new "experimental" developments go here, as do submissions from new authors, then periodically when these are all looking stable they get merged to Trunk for full testing, and then when that's all looking OK, merged to the release branch for release. In contrast bug fixes to existing code, generally go straight into the Trunk so that users can use these right away, and likewise they tested in our test matrix right away. These get merged back into the sandbox development branch at the same time as sandbox changes are merged into Trunk. But anyhow, that's digressing from the question :-) HTH, John.