trickle.py is broken #10208

Activity

1. annevk

   added

   xhr

   fetch

   on Mar 28, 2018
2. jgraham commented on Mar 28, 2018

   

   jgraham

   on Mar 28, 2018

   Contributor

   More actions

   FWIW you might be better off using the trickle pipe i.e. load a resource with something like ?pipes=trickle(10:d2:r2) meaning "Send 10 bytes, wait 2s, repeat the previous two steps until the resource is fully sent"
3. annevk commented on Mar 28, 2018

   

   annevk

   on Mar 28, 2018

   MemberAuthor

   More actions

   That might be interesting. I also now wonder if what I observed was instead an HTML parser thingy and not a server thingy. Where the HTML parser wants at least a 1000 bytes or so.
4. jgraham commented on Mar 28, 2018

   

   jgraham

   on Mar 28, 2018

   Contributor

   More actions

   Yeah, it looks like flush() is called implicitly unless you disabled that, so I think the code as written should work.
5. zcorpan commented on Apr 19, 2018

   

   zcorpan

   on Apr 19, 2018

   Member

   More actions

   So this bug affects these 11 tests:

   $ grep -r "trickle.py" .
   ./xhr/firing-events-http-content-length.html:        xhr.open("GET", "resources/trickle.py?ms=0&count=100&specifylength=1", true);
   ./xhr/firing-events-http-no-content-length.html:        xhr.open("GET", "resources/trickle.py?ms=0&count=100", true);
   ./xhr/send-timeout-events.htm:            xhr.open("POST", "./resources/trickle.py", true);
   ./xhr/response-data-progress.htm:async_test(function () { doTest(this, false, 0, "resources/trickle.py?count=6&delay=150"); },
   ./xhr/response-data-progress.htm:async_test(function () { doTest(this, true, 78, "resources/trickle.py?count=6&delay=150&specifylength=1"); },
   ./xhr/event-loadstart-upload.htm:    client.open("POST", "resources/trickle.py?ms=5&count=8");
   ./xhr/event-readystatechange-loaded.htm:    client.open("GET", "resources/trickle.py?count=10"); // default timeout in trickle.py is 1/2 sec, so this request will take 5 seconds to complete
   ./xhr/event-progress.htm:    client.open("GET", "resources/trickle.py?count=4&delay=150");
   ./fetch/api/response/response-cancel-stream.html:    return fetch(RESOURCES_DIR + "trickle.py?ms=30&count=100").then(function(response) {
   ./fetch/api/response/response-cancel-stream.html:    return fetch(RESOURCES_DIR + "trickle.py?ms=30&count=100").then(function(response) {
   ./fetch/api/response/response-clone.html:        return fetch('../resources/trickle.py?count=2&delay=100').then(function(res) {
   ./fetch/api/response/response-clone.html:        return fetch('../resources/trickle.py?count=2&delay=100').then(function(res) {
   ./fetch/api/basic/stream-response.any.js:  return fetch(RESOURCES_DIR + "trickle.py?ms=30&count=100").then(function(resp) {
   ./fetch/api/request/request-keepalive-quota.html:            // the effects of a rolling quota. To do this we will utilize trickle.py with a 1s delay. This should
   ./fetch/api/request/request-keepalive-quota.html:            const trickleURL = '../resources/trickle.py?count=1&ms=';
   

   The HTML parser waiting for 1024 bytes or encoding declaration sounds plausible but I haven't tested to confirm that is what is happening here...
6. zcorpan commented on Apr 19, 2018

   

   zcorpan

   on Apr 19, 2018

   Member

   More actions

   On irc

   wanderview zcorpan: while trying to get around the parser wait in firefox I also had to make sure the response had a Content-Type header... not sure if that impacts you
   wanderview Content-Type: text/html
   
7. annevk commented on Apr 24, 2018

   

   annevk

   on Apr 24, 2018

   MemberAuthor

   More actions

   Adding ;charset=UTF-8 definitely changes things in Firefox when loading this in a browsing context, but Chrome doesn't seem to be affected and doesn't display anything faster.

   It probably doesn't affect any of the 11 tests above though, even though they could use some improvements...
8. annevk commented on Apr 24, 2018

   

   annevk

   on Apr 24, 2018

   MemberAuthor

   More actions

   It seems to affect Chrome for xhr/response-data-progress.htm.