- Last edited November 28, 2002 |
The only thing I can find on-line is (btw, the email address for IvanMoore has changed since then):
ciclamino.dibe.unige.it/XP2000/summaries/GammaSummary.htmIBM's eclipse has streams which are based on the same model, but are implemented on top of CVS:
www.eclipse.org/documentation/html/plugins/org.eclipse.platform.doc.user/doc/reference/ref-51.htm---
I've used the streams in eclipse for a while and they do an excellent job of simplifying the age old job of catch up + merge. In fact streams are so good they almost entirely hide the fact that you are using a version control system from the dark ages (CVS).
However, Ivan has convinced me that the Eclipse streams are not as good as team streams in java micro edition. I guess this comes down to the underlying limitations of CVS - there is no way to store meta data relating to an atomic release of many updated files. So rolling back after you have performed a release is erm - non-trivial. MpC
Would this be sorted out if Eclipse implemented streams on top of Subversion instead?
I doubt it. As I attempted to describe in SimpleVersionControl, existing file-based source-code systems are a very poor match for todays refactoring-intensive, project-based, fast-iterating development. Just about every developer needs some sort of "one button" resolve-conflicts/release or rollback process, and it seems crazy to have to manually tell CVS (and subversion!) that you have added a file to a project -- you don't have to tell it you have added a line to a file, and I don't see any important conceptual difference. And you just try to remove or rename a file in either! Sure, subversion preserves a bit more metadata, but it still doesn't solve the basic problem.
Look at a list of refactorings one day and see how many of them need you to manually (and dangerously) mess with your version-control system. If you can find a VC system which transparently copes with all the refactorings, then please tell us all --FrankCarver
- Last edited November 28, 2002 |