[Varnish] #411: restart not available in vcl_deliver

[Varnish]

#411: restart not available in vcl_deliver
----------------------+-----------------------------------------------------
 Reporter:  hajile    |       Owner:  phk  
     Type:  defect    |      Status:  new  
 Priority:  normal    |   Milestone:       
Component:  varnishd  |     Version:  trunk
 Severity:  normal    |    Keywords:       
----------------------+-----------------------------------------------------
 Reason for restart in vcl_deliver (not in fetch or hit) is to cache a
 redirect (in my case 303: See other) in fetch phase, then restart with
 url=Location right in varnish, without redirecting the client (thus
 optimizing internal network traffic).
 Both operations are perfectly cacheable and at the same time consistent
 (no need to purge two urls, purge only the primary one).

 This feature is very useful for proxying RESTful services:
 /object/{id} -> return document [cached, purge when return document
 changes]
 /object/?findByNonPrimaryButUniqueKey=bla -> HTTP 303 ( /object/{id} )
 [cached, purge only when this unique key changes]
 /object/?findByOtherNonPrimaryButUniqueKey=22895723658 -> HTTP 303 (
 /object/{id} ) [cached, -""-]

 sub vcl_hit {
  if (obj.http.X-Magic-Redirect == "1") {
   set req.url = obj.http.Location;
   restart;
  }
 }

 sub vcl_fetch {
  if (obj.status == 301 || obj.status == 302 || obj.status == 303 ||
 obj.status == 307) {
   if (obj.status == 303) {
    set obj.http.X-Magic-Redirect = "1";
   }
   deliver;
  }
 }

 sub vcl_deliver {
  if (resp.http.X-Magic-Redirect == "1") {
   unset resp.http.X-Magic-Redirect;
   restart;
  }
  deliver;
 }

-- 
Ticket URL: <http://varnish.projects.linpro.no/ticket/411>
Varnish <http://varnish.projects.linpro.no/>
The Varnish HTTP Accelerator