| 1 | | I agree with you about the second option after researching the RFC. But the first option does not make sense to me, and I cannot find anything in the RFC to support it. |
| | 1 | After researching the RFC, I agree with you about the second option. |
| | 2 | |
| | 3 | But the first option does not make sense to me, and I cannot find anything in the RFC to support it. |
| | 4 | |
| | 5 | Assuming closing the connection is the only sensible thing to do, then the Amazon ELB is definitely behaving poorly by offering the 'connection: keep-alive' in its 504 response. And this seems to be a knock on effect because of the nginx TCP connection drop. |
