CVSup is a software package for distributing and updating collections of files across a network. It can efficiently and accurately mirror all types of files, including sources, binaries, hard links, symbolic links, and even device nodes. CVSup’s streaming communication protocol and multithreaded architecture make it most likely the fastest mirroring tool in existence today.
CVSup was conceived and developed by John Polstra, a consultant whose specialties include high-performance network communications.
In addition to being a great general-purpose mirroring tool, CVSup includes special features and optimizations specifically tailored to CVS repositories. By taking advantage of the special properties of the files contained in CVS repositories, CVSup is able to perform updates much faster than traditional systems. CVSup parses and understands the RCS files making up a CVS repository. When updates occur, CVSup extracts new deltas directly from the RCS files on the server and edits them into the client’s RCS files. Likewise, CVSup notes the addition of new symbolic tags to the files on the server and sends only the new tags to the client.
CVSup is able to merge new deltas and tags from the server with deltas and tags added locally on the client machine. This makes it possible for the client to check local modifications into his repository without their being obliterated by subsequent updates from the server.
In addition to distributing the RCS files themselves, CVSup is able to distribute specific checked-out versions. The client can specify a symbolic tag, a date, or both and CVSup will extract the appropriate versions from the server’s CVS repository. Checked-out versions do not need to be stored on the server, since CVSup can extract any version directly from the CVS repository. If the client has an existing checked-out tree, CVSup will apply the appropriate edits to update the tree or transform it into the requested version. Only the differences between the existing version and the desired version are sent across the network.
To update non-RCS files, CVSup uses the highly efficient rsync algorithm, developed by Andrew Tridgell and Paul Mackerras.
CVSup uses lightweight processes (threads) to implement a streaming protocol across the network. This completely eliminates the delays associated with the lock-step, request-reply form of communication used by many existing protocols, such as sup and NNTP. Information is transferred at the full available speed of the network in both directions at once. Network latency and server response delays are rendered practically irrelevant.
CVSup uses the “zlib” compression package to optionally compress all communications. This provides an additional 65-75% compression, on top of the diff-based compression already built into CVSup.