Index: dbfile.scm
==================================================================
--- dbfile.scm
+++ dbfile.scm
@@ -528,17 +528,17 @@
 			    ))))))
 	 (on-tmp      (equal? (car (string-split dbpath "/")) "tmp"))
 	 (db        (if on-tmp
 			(dbfile:cautious-open-database dbname init-proc 1 "WAL" force-init: #t) ;; WAL MODE should use syncronous=1
 			;; (dbfile:cautious-open-database dbname init-proc 0 #f    force-init: #t)
-			(dbfile:cautious-open-database dbname init-proc 0 "MEMORY" force-init: #t)
+			(dbfile:cautious-open-database dbname init-proc 0 "MEMORY" force-init: #t) ;; Journal mode = memory is fastest?
 			;; (sqlite3:open-database dbname)
 			)))
-    (if on-tmp	      ;; done in cautious-open-database
-	(begin
-	  ;; (sqlite3:execute db "PRAGMA synchronous = 0;") ;; why was this here when is is handled by cautious-open-database?
-	  (sqlite3:set-busy-handler! db (sqlite3:make-busy-timeout 136000))))
+    ;; (if on-tmp	      ;; done in cautious-open-database
+    ;;    (begin
+    ;; (sqlite3:execute db "PRAGMA synchronous = 0;") ;; why was this here when is is handled by cautious-open-database?
+    (sqlite3:set-busy-handler! db (sqlite3:make-busy-timeout 136000)) ;; ))
     db))
 
 ;; mtest processes registry calls
 
 (define (dbfile:insert-or-update-process nsdb dat)
@@ -582,18 +582,20 @@
 		   host port pid starttime endtime status purpose dbname mtversion))
 
 (define (dbfile:set-process-status nsdb host pid newstatus)
   (sqlite3:execute nsdb "UPDATE processes SET status=? WHERE host=? AND pid=?;" newstatus host pid))
 
+;; as sorted should be stable. can use to choose "winner"
+;;
 (define (dbfile:get-process-options nsdb purpose dbname)
   (sqlite3:fold-row
    ;; host port pid starttime status mtversion
    (lambda (res . row)
      (cons row res))
    '()
    nsdb
-   "SELECT host,port,pid,starttime,endtime,status,mtversion FROM processes WHERE purpose=? AND dbname LIKE ? AND status='alive';"
+   "SELECT host,port,pid,starttime,endtime,status,mtversion FROM processes WHERE purpose=? AND dbname LIKE ? AND status IN ('running','alive') ORDER BY starttime ASC,host,port;"
    purpose dbname))
 
 (define (dbfile:get-process-info nsdb host pid)
   (let ((res (sqlite3:fold-row
 	      ;; host port pid starttime status mtversion
@@ -604,17 +606,25 @@
 	      "SELECT host,port,pid,starttime,endtime,status,purpose,dbname,mtversion FROM processes WHERE host=? AND pid=?;"
 	      host pid)))
     (if (null? res)
 	#f
 	(car res))))
+
+(define (dbfile:row->procinf row)
+  (match row
+   ((host port pid starttime endtime status mtversion)
+    (make-procinf host: host port: port pid: pid starttime: starttime endtime: endtime status: status mtversion: mtversion))
+   (else
+    (debug:print 0 *default-log-port* "ERROR: row "row" did not match host,port,pid,starttime,endtime,status,mtversion")
+    #f)))
 
 (define (dbfile:set-process-done nsdb host pid reason)
-  (sqlite3:execute nsdb "UPDATE processes SET status='ended',endtime=?,reason=? WHERE host=? AND pid=?;" (current-seconds) reason host pid)
+  (sqlite3:execute nsdb "UPDATE processes SET status='done',endtime=?,reason=? WHERE host=? AND pid=?;" (current-seconds) reason host pid)
   (dbfile:cleanup-old-entries nsdb))
 
 (define (dbfile:cleanup-old-entries nsdb)
-  (sqlite3:execute nsdb "DELETE FROM process WHERE status='ended' AND endtime<?;" (- (current-seconds) (* 3600 48))))
+  (sqlite3:execute nsdb "DELETE FROM process WHERE status='done' AND endtime<?;" (- (current-seconds) (* 3600 48))))
 
 ;; other no-sync functions
 
 (define (dbfile:with-no-sync-db dbpath proc)
   (mutex-lock! *no-sync-db-mutex*)

Index: tcp-transportmod.scm
==================================================================
--- tcp-transportmod.scm
+++ tcp-transportmod.scm
@@ -564,11 +564,35 @@
 		       ((null? servers) #f) ;; not ok
 		       ((equal? (list-ref (car servers) 6) ;; compare the servinfofile
 				(tt-servinf-file ttdat))
 			(let* ((res (if db-locked-in
 					#t
-					(let* ((lock-result  ;; this is the primary lock - need to double verify that got it
+					;;
+					;; let's replace the below "winning" lock method with:
+					;;  1. create a lock file with pid etc.
+					;;  2. if there are no other lock files make an entry in the no-sync db
+					;;  3. gather the lock entries, apply the "winner" heuristic
+					;;  4. if I'm the winner, set tt-state to 'running else set to 'notthewinner
+					;;
+					;; New idea:
+					;;  1. check all processes entries that match the db
+					;;  2. sort by fixed heuristic
+					;;  3. if I'm number one, set state to 'running and db-locked-in to #t
+					(let* ((candidates (dbfile:with-no-sync-db
+							     nosyncdbpath
+							     (lambda (nsdb)
+							       (dbfile:get-process-options nsdb "server" dbfname))))
+					       (primecand  (begin
+							     (assert (not (null? candidates))
+								     "HOW CAN WE NOT BE IN THE PROCESSES DB AS A SERVER?")
+							     (dbfile:row->procinf (car candidates)))))
+					  ;; compare primecand with myself
+					  ;; if not me check that it is reachable
+					  ;;   if reachable exit
+					  #f)
+					
+					#;(let* ((lock-result  ;; this is the primary lock - need to double verify that got it
 						(dbfile:with-no-sync-db
 						 nosyncdbpath
 						 (lambda (db)
 						   (db:no-sync-lock-and-check db dbfname
 									      (tt-servinf-file ttdat)