| 11 Sep 2021 |
 baloo | I think it would make sense to juste use max-age and not s-maxage | 15:54:44 |
| baloo | * a max-age in the Cache-Control header would serve the same purpose, and allow each layer to implement the same strategy. | 15:55:11 |
| 12 Sep 2021 |
 Jonas Chevalier | yeah that would work with Fastly as well. One thing I noticed though is that Fastly doesn't forward the Cache-Control header, not sure why. | 10:08:52 |
| Jonas Chevalier | HTTP/2 301
location: https://releases.nixos.org/nixos/21.05/nixos-21.05.3021.8b0b81dab17
server: AmazonS3
via: 1.1 varnish, 1.1 varnish
access-control-allow-origin: *
accept-ranges: bytes
date: Sun, 12 Sep 2021 10:07:44 GMT
age: 69302
x-served-by: cache-bwi5126-BWI, cache-cdg20759-CDG
x-cache: HIT, HIT
x-cache-hits: 1, 1
x-timer: S1631441264.264690,VS0,VE2
content-length: 0
| 10:09:05 |
| Jonas Chevalier | also 301 should be changed to 302 since it's not permanent | 10:09:27 |
| Jonas Chevalier | deleting route53 references since we're using the netlify DNS | 10:14:02 |
 Sandro | Does the cache control Header even matter? Does nix look at it? Maybe the caching issues is only present in the browser? | 10:15:03 |
| Jonas Chevalier | most CLI tools don't have a cache and so won't act on the headers | 10:16:05 |
| Sandro | That's what I am thinking, too | 10:17:06 |
| Jonas Chevalier | If we follow the HTTP protocol more closely, it will be better for us in the long run. | 10:18:23 |
| Jonas Chevalier | I have seen that happen many times where clients then start implementing workarounds, and then it becomes hard to fix the non-conforming API while keeping back-compat. | 10:19:06 |
| Jonas Chevalier | * deleting route53 references since we're using the netlify DNS (see https://github.com/NixOS/nixos-org-configurations/pull/174) | 10:20:54 |
 sterni | In reply to @sandro:supersandro.de
Does the cache control Header even matter? Does nix look at it? Maybe the caching issues is only present in the browser? the issue is fastly caching server side not some tool acting on the cache-control header client side | 12:00:12 |
 Vladimír Čunát | I suspect such things might still become relevant in some cases, e.g. caching proxies. (Another layer. I think I've read about some Chinese group doing something like that due to our cache being too slow in there.) | 12:55:10 |
|  MarkN joined the room. | 20:39:02 |
| baloo | last time I researched for work (~3y ago), getting a cdn in china was ... challenging | 20:40:19 |
| baloo | at least, not without giving a tls key to an unknown entity | 20:40:45 |
| baloo | cloudflare has (had?) one, but that was a contractor handling traffic in china | 20:41:34 |
| 13 Sep 2021 |
| Vladimír Čunát | Ever seen this on a hydra?
Error pulling from the bitbucket pulls API: {"type": "error", "error": {"message": "The requested repository either does not exist or you do not have access. If you believe this repository exists and you have access, make sure you're authenticated."}}
| 09:23:39 |
| Vladimír Čunát | (of course, the repo+branch is unchanged - the normal public nixpkgs) | 09:24:15 |
| Vladimír Čunát | * (of course, the repo+branch are unchanged - the normal public nixpkgs) | 09:24:25 |
 toonn | There's a mirror on Bitbucket? | 09:53:17 |
| Vladimír Čunát | The config doesn't attempt to do anything with bitbucket. | 10:03:24 |
| toonn | Oh, bizarre. | 10:07:57 |
| Jonas Chevalier | 
Download image.png | 10:29:58 |
| Jonas Chevalier | why is the supportedSystems type "open bitbucket pull requests" | 10:30:19 |
 lukegb (he/him) | Hah | 10:32:53 |
| lukegb (he/him) | Terraform? | 10:33:00 |
| Jonas Chevalier | I restored it to "Nix expression" by hand, let me know if it drifts back | 10:33:33 |
| Vladimír Čunát | Oh, my bad. I was being an idiot. | 10:39:24 |
