Brian explains.

diff --git a/sccs2rcs b/sccs2rcs
index 4de7cbb68f276476574caf13720f80eafeb382da..6752430b5a0648eb6eb88ddd17abb72dcb742fd2 100755 (executable)
--- a/sccs2rcs
+++ b/sccs2rcs
@@ -9,10 +9,10 @@
 # In 1993 I applied some minor fixes for newer RCS versions and it 
 # shipped as part of one of Rich Morin's Prime Time Freeware releases.
 #
-# At some undetermined time after 1993, Brian Berliner hacked Ken
-# Cox's original in a slightly different direction than mine, adding
-# more SCCS keyword translations and an improvement in dealing with 
-# branches.
+# In 1994 Brian Berliner hacked Ken Cox's original in a slightly
+# different direction than mine, adding more SCCS keyword translations
+# and an improvement in dealing with branches.  At least, Brian *thinks*
+# it was 1994...
 #
 # In 2002 I fixed it not to require the Berkeley sccs(1) front end.
 # In 2003 I folded in a minor bug fix from Roger Pilkey.
@@ -290,9 +290,15 @@ if __name__ == '__main__':
                 lckrev = ".".join(revparts[:-1])
                 if rev.count(".") > 1:
                     # need to lock the branch -- it is OK if the lock fails
-                    # Above comment was from Brian Berliner's csh version,
-                    # but it's not at all clear why this lock can be allowed
-                    # to fail while the trunk one must succeed.
+                    #
+                    # I asked Brian and he explained this as follows:
+                    # "As it turns out, with RCS, if you are adding the
+                    # first revision of a new branch, you don't need
+                    # to lock the trunk. In fact, you don't want to
+                    # lock the trunk because checking in the revision
+                    # on the branch will not unlock the trunk. So, the
+                    # hack I applied was to get it to add branches
+                    # correctly."
                     do_it("rcs -l%s %s" % (lckrev, stem))
                     record("got branch lock")
                 else: