Merged "be subscribe" and be-handle-mail subscription support.

Also assorted other changes and fixes in the be.subscribe branch.
Highlights:
  * Much more powerful libbe.diff with subclassable report generators.
  * "be diff" compares working copy with last commit by default.
  * comment reference text shown in "be comment" EDITOR footer
  * .revision_id() for all VCSs
  * meaningful comment comparison and stricter bug comparison
  * stricter .sync_with_disk interpretation.  See BugDir.__doc__.
  * Comment.From and .time_string -> .author and .date, for better
    conformance with settings_object.setting_name_to_attr_name().

1 files changed, 20 insertions, 0 deletions

diff --git a/.be/bugs/52034fd0-ec50-424d-b25d-2beaf2d2c317/comments/4c50ca0b-a08f-4723-b00d-4bf342cf86b6/body b/.be/bugs/52034fd0-ec50-424d-b25d-2beaf2d2c317/comments/4c50ca0b-a08f-4723-b00d-4bf342cf86b6/body
new file mode 100644
index 0000000..90b386a
--- /dev/null
+++ b/.be/bugs/52034fd0-ec50-424d-b25d-2beaf2d2c317/comments/4c50ca0b-a08f-4723-b00d-4bf342cf86b6/body
@@ -0,0 +1,20 @@
+I'm all for flexibility, so long as it doesn't require too much
+hackery to implement it.  You'll have two problems:
+
+  * Determining what to commit.
+
+    You'd have to have RCS keep a log of all versioned files it
+    touched, and extend .commit() to accept the keyword list "files"
+    and commit only those files.  This is doable, but maybe not worth
+    the trouble.
+
+  * Generating meaningful commit messages.
+
+    You'd have to add this functionality to each command (and future
+    commands).
+
+This would probably not be a good idea for the Arch and Mercurial
+backends, since they have a limited ability to rewrite history when
+you screw up your commit message (as far as I can tell).  Mercurial
+does have "hg rollback", but it only works once, and lots of
+typo-correction commits would just make the logs awkward.