diff --git a/doc/bugs/git-annex_sync_under_windows_fails_by_using_itself_as_the_ssh_command/comment_11_28bfd4d05068748fab2daf1baa2a83cc._comment b/doc/bugs/git-annex_sync_under_windows_fails_by_using_itself_as_the_ssh_command/comment_11_28bfd4d05068748fab2daf1baa2a83cc._comment
new file mode 100644
index 0000000000..8772b7307d
--- /dev/null
+++ b/doc/bugs/git-annex_sync_under_windows_fails_by_using_itself_as_the_ssh_command/comment_11_28bfd4d05068748fab2daf1baa2a83cc._comment
@@ -0,0 +1,22 @@
+[[!comment format=mdwn
+ username="joey"
+ subject="""comment 11"""
+ date="2015-08-04T19:03:54Z"
+ content="""
+I took another look at this, and I have a suspicion that the difference is
+window's behavior when an env var is set to the empty string.
+
+On Unix, that has a distinct meaning from an env var being set. I suspect
+that it doesn't on Windows. Or perhaps on some versions of windows and/or
+some versions of haskell libraries on some versions of windows... (Since I
+still can't reproduce the problem on windows myself.)
+
+If `GIT_ANNEX_SSHOPTION=""` results in it not being set, then the rest
+of this bug report follows.
+
+I've made a simple change. If the option is empty, it doesn't try to set
+`GIT_SSH` at all. I am pretty sure this entirely fixes the problem.
+
+Please test the windows daily build, which has this fix applied to it,
+and get back to me if not.
+"""]]