2010-10-11 09:37 AEST by Arthur Barrett - RCS revisions should have a sequence and should be able to select a sequence 
using -r 999.SEQ

This will produce a SVN like result. eg:

  cvs co -r 999.23 module 

Will select revisions where sequence is 23 or lower for all files in module. 
Not dissimilar to commitid - but where commitid selects only that one (set of) 
revision, the sequence is used for a range (that set of revisions and all 
earlier ones). 

The prefix 999 should be user definable (unlike @commitid). 

Combined Cvsmanager & cvslock needs an sqlite database. Server process obtains 
a new sequence number when it creates a new revision. If RCS file cannot be 
renamed then we end up with a gap in sequence but that is OK. 

Does not rely on dates - just look for the same places that were modified to 
support commit id's and add the code to support commit sequence.

2011-02-06 07:43 AEST by Arthur Barrett - Why do this?

The reason(s) for this are:
* equivalent to 'auto-tagging' repo for each commit 
* easily select a range of revisions (that set of revisions and all earlier ones). 
* improve update speed (is repo rev higher than last update rev?)


How should branches be distinguished?  Eg: there could be a new commit to branch 'abc' but a sandbox is 
for branch 'efg' - so is the fact that the revision number is higher matter?