Each time a "child" stream promotes a new baseline, it dynamically updates the "current configuration" of the composite codeline to refer to the latest baselines of the child component baselines. See references on Component/Container-Based SCM 
Task/Activity-Based Development (TBD) is already popular in many high-end version control tools. Yet developers more accustomed to lowest-common-denominator tools are less familiar with this feature and approximate it (often manually) via Task-Level Commit (TLC) and Change-Sets or Change-Packages. We predict that use of TLC will evolve into full-fledged TBD  as integration with project-management and change tracking tools becomes more prevalent in agile development shops.
Continuous Update  will ride along the coattails of TBD (along with finer-grained change tasks) and more tools will boast either virtual-file-system capabilities or else provide both "push" and "pull" synchronization mechanisms for updating private workspaces.
Improved Refactoring Support for renaming and splitting/combining files and directories/folders, along with directory versioning will also become a more commonly supported feature. Also look for language/syntax-aware merging  to make its way into popular IDEs (e.g. Eclipse).
Trends toward "Agile" Build/Test Management
To many, build management is the "bread-and-butter" of CM. Look for trends and improvements in build management architecture for Agile projects.
Continuous Integration "Architecture" - look for individual projects and teams to avail themselves of all sorts of technology and tools to improve the frequency, reliability/quality/stability, and performance of both full and incremental builds. Asynchronous integration+build+test may even call upon technologies such as grid-computing, autonomic-computing, and on-demand computing to leverage and load-balance servers for building and automated testing. Expect a combination of open-source and commercial tools to be utilized.
Enterprise Continuous Integration/Staging - look for multi-group coordination of multiple component-teams to enhance their "build/integration architectures" to include dynamic event-triggered assembling and re-building of systems and subsystems from recently built+tested component versions utilizing multiple staging/integration areas (possibly distributed or replicated)
Testing Framework Integration - the agile/XP focus on test-driven development will drive a demand for better integration between SCM tools and not just unit-test automation tools (e.g., JUnit) but also testing frameworks. IBM/Rational already has a jump on this with the Hyades framework within Eclipse .
Trends toward "Agile" Distributed/Collaborative Development
As mentioned by some of our fellow columnists, the growth of "offshoring" and other forms of distributed development is making this aspect of CM tools much more important. How can we manage remote teams working on the same source base?
Will distributed-patch-based version-control paradigms such as that of Bitkeeper and Arch, where each workspace is a first-class repository, become more in-demand and more prevalent? If so, then this would make the need for things like private-branches and task-branches disappear (because you wouldn't need to "branch-in-advance of doing a commit"). The alternative approach is the stream-based model used by ClearCase/UCM, Accu-Rev, Subversion, Perforce, SpectrumSCM, OurayCM, and others. See David Wheeler's essay on centralized -vs- decentralized SCM systems .
The distributed model has obvious performance advantages. However, corporate clients often prefer to have their repository managed centrally, despite of the potential performance implications. It can be much more secure and easier to ensure backups are in place ? they don't like the thought of software "inventory" (work-in-progress) sitting on desktops or laptops and not well managed. In an interesting take on this, Joel Spolsky notes :
What's kind of surprising is that it has turned out to be easier to rewire the entire world for high-bandwidth Internet than it is to create good replication architecture so you can work disconnected!
Regardless of which way the trend goes (or whether it becomes a melding of the two),