Extend sampleconf, explain what stdinserver serve type does.

cmd/sampleconf/localbackup/host1.yml

@@ -10,6 +10,7 @@ mirror_local:
   }
   snapshot_prefix: zrepl_
   interval: 10m
+  initial_repl_policy: most_recent
 
   # keep one hour of 10m interval snapshots of filesystems matched by
   # the left-hand-side of the mapping

cmd/sampleconf/pullbackup/backuphost.yml

@@ -13,8 +13,10 @@ fullbackup_prod1:
   mapping: {
     "<":"storage/backups/zrepl/pull/prod1.example.com"
   }
+  initial_repl_policy: most_recent
 
   # follow a grandfathering scheme for filesystems on the right-hand-side of the mapping
+  snapshot_prefix: zrepl_
   prune:
     policy: grid
     grid: 1x1h(keep=all) | 24x1h | 35x1d | 6x30d

cmd/sampleconf/pullbackup/productionhost.yml

@@ -1,8 +1,13 @@
 fullbackup_prod1:
 
   # expect remote to connect via ssh+stdinserver with fullbackup_prod1 as client_identity
-  type: pull-source
+  type: source
   serve:
+    # Creates an AF_UNIX socket with name client_identity in a well-known directory
+    # private to the zrepl user (which runs both the zrepld and the stdinserver command via authorized_keys)
+    # The stdinserver command connects to that socket and sends its stdin and stdout
+    # file descriptors over that UNIX socket to the zrepld.
+    # This avoids additional memory-to-memory copies and is more portable than splice(2) on Linux.
     type: stdinserver
     client_identity: fullbackup_prod1
 
@@ -14,6 +19,7 @@ fullbackup_prod1:
   snapshot_prefix: zrepl_
   interval: 10m
 
+
   # keep a one day window 10m interval snapshots in case pull doesn't work (link down, etc)
   # (we cannot keep more than one day because this host will run out of disk space)
   prune: