| 19 Feb 2022 |
 f0x | maybe other parts changed though | 01:12:33 |
| f0x | using a hash of the authorization header could work too? | 01:12:52 |
 Dandellion | huh you're right | 01:13:11 |
 hexa | yeah, so how do you access that in nginx? | 01:16:07 |
| f0x | $http_Authorization apparently | 01:17:51 |
| hexa | neat | 01:17:58 |
| hexa | so it's just hash $http_Authorization and we'd be good? | 01:18:30 |
| f0x | i have no idea but it's worth a try :P | 01:18:55 |
| hexa | so I wonder, you're saying that ideally we'd support both | 01:21:05 |
| hexa | but what would you people actually be using if both were available? | 01:21:16 |
| Dandellion | since the same data needs to be synced to all devices, currently it makes more sense let the workers cache things by username and not device | 01:22:07 |
| Dandellion | I don't think it actually matters for anything but the largest installations | 01:22:29 |
| hexa | so we have two means to get the user, by IP or by Authorization header | 01:22:45 |
| hexa | * so we have two means to get the device, by IP or by Authorization header | 01:22:53 |
| f0x | In reply to @hexa:lossy.network
but what would you people actually be using if both were available? the vast majority just needs some fixed worker types | 01:23:02 |
| Dandellion | i'd use IP in that case | 01:23:02 |
| hexa | but how would we generalize towards the user? | 01:23:04 |
| f0x | to split the important parts into a few processes | 01:23:11 |
| f0x | tbqh most people might not even need multiple of the difficult to load balance types | 01:23:52 |
| hexa | anyway, if you have ideas for how to solve these problems, dump them into the document | 01:27:18 |
| hexa | I'm off for the night \o | 01:27:31 |
| Dandellion | It's 2 am for me as well, I'll share my thoughts tomorrow :) | 01:28:06 |
| 21 Feb 2022 |
|  continuouswave changed their display name from Chuck Winter to Chinchilla Wetreat. | 00:47:14 |
|  @janejasperous:one.ems.host joined the room. | 04:28:35 |
| hexa | ma27: what's your schedule? 🙂 | 23:00:00 |
| 22 Feb 2022 |
| hexa | added two ideas on how to delay worker starts to the pad | 16:03:43 |
| hexa | one uses systemctl is-active, the other uses the /health endpoint on the main instance | 16:04:18 |
| hexa | Also Dandellion's worker units should probably use Requires=matrix-synapse.service, so the workers shut down when th main instance does | 16:07:03 |
| hexa | * Also Dandellion's worker units should probably use Requires=matrix-synapse.service, so the workers shut down when the main instance does | 16:07:13 |
| Dandellion | I'll add it to the requires if enableMainSynapse is true ig | 17:32:41 |
