WebDriver/Meetings/2018-06-25: Difference between revisions

WebDriver/Meetings/2018-06-25 (view source)

Revision as of 14:11, 25 June 2018

1,581 bytes added , 25 June 2018

→‎HTTP/1.1 Keep-Alive connections: add minutes

Ato

715

edits

@@ Line 21: / Line 21: @@
 ; ato
-: Said something.
+: Not sure what’s going on, but we have three fresh issues on GitHub that could be affected by the <code>Keep-Alive</code> change.
-: And more things.
+; whimboo
+: Not happy this landed before 0.21.0, we might have to revert it and release a patch.
+: Maybe gate it on a capability?
+; ato
+: Need to be set as HTTPD starts, so capability not a possibility.
+; whimboo
+: Where does the wait happen? In client or in server?
+; ato
+: My impression was that it happened in the client.  They wait for ''N'' seconds, then after that the next command fails.
+: Question is, do we think this is caused by <code>Keep-Alive</code>?
+; whimboo
+: I would say so.
+; ato
+: Do we then back the change out and do a patch release?
+; whimboo
+: How many use this by default?
+; ato
+: If the client asks for it, the server provides I believe.  There is a negotiation between the client and the server about capabilities.
+: It’s interesting that this only appears to happen in the Python client?
+; whimboo
+: From the trace-log I don’t see that the connection was closed.
+: Which versino of Python? Maybe only Python 3 related?
+; ato
+: There’s a lot of things to debug here.
+: Do we back out of leave it to stir another couple of days?
+; whimboo
+: Maybe leave it a little while longer.
+; ato
+: Yes, I suppose we don’t want to have to do a lot of work if it only impacts small number of people.
+; whimboo
+: Yes, let’s leave it and discuss with David via email
+Conclusions:
+* Leave patch in for the moment, observe how it affects users
+* Discuss with AutomatedTester
+* Triage on geckodriver issues, possibly dupe against old <code>Keep-Alive</code> request
 === Status updates ===