| 28 Oct 2021 |
 Domen Kožar | It also holds us back, as we could use foundation funds for things like events, more infrastructure, etc | 00:10:49 |
| Domen Kožar | And as mentioned above, it's not transparent as there's almost no expenses, so why would the companies donate? | 00:12:46 |
| Domen Kožar | My motivation why it's important to prioritize and fix this properly. | 00:13:23 |
| Domen Kožar | * A few thoughts why it's important to prioritize and fix this properly. | 00:13:43 |
| Domen Kožar | * And as mentioned above, it's not transparent as there're almost no expenses, so why would the companies donate? | 00:25:42 |
 tomberek | The "risk expense" is not adequately captured in the reporting of expenses. | 00:36:36 |
 sterni | if we really have to almost 90% of everything raised via opencollective as risk contingency that sounds like a very bad place to be in | 09:30:19 |
| sterni | * if we really have to almost 90% of everything raised via opencollective as risk contingency that sounds like a suboptimal place to be in | 09:34:06 |
|  ma27 joined the room. | 10:07:32 |
 andi- | People hate the topic but if the funds we would need are really in the 10k/month region we should perhaps think of owning the hardware and having just a caching CDN. Still lots of work (hardware I know...) but perhaps that wouldn't be the end of the world if it means that we don't have to fear a random sponsor leaving the ship. | 10:55:35 |
| sterni | I think our inability to maintain the infrastructure we do own doesn't bode well for such an undertaking :p | 11:00:15 |
| andi- | I know. It would probably not work with the current setup of infra team. | 11:04:55 |
| andi- | But it would mean a lot less than 10k/m of opex for a larger (30k ish?) one time investment and smaller monthly fees (hosting, bandwidth, replacement hardware, ..) | 11:05:41 |
 Sandro | Buying hardware currently is not that easy. There are usually delays of months. | 11:22:09 |
 Alyssa Ross | there's not huge time pressure | 11:22:43 |
| Sandro | If you can buy something at all. | 11:25:57 |
| Sandro | If we host our own hardware we would either need to increase the storage once in a while or start pruning things | 11:26:40 |
| Sandro | Also minio the only S3 server I know of is licensed under AGPL | 11:27:06 |
 Jamie | that presumably wouldn't be a problem for us? | 11:28:44 |
| Jamie | unless our infra team is all google employees anyway | 11:29:31 |
| Jamie | in which case we should clearly be using their wonderful AMP proxies to serve our files /s | 11:30:13 |
| Sandro | Also doing caching in front of that is also not that easy because we need at least locations in US and EU.
If we do US only the cache is horrible slow in Europe especially Germany because ISPs have terrible peering here. I tested that recently with cachix and the slowdown was over 20x on gigabit. | 11:31:16 |
| Sandro | Also what happens if the cache has problems and is down? Can we guarantee it will be fixed in X hours or will it take days? | 11:32:16 |
| Jamie | my reading of the suggesting is that our hardware would just be the backing server for another caching cdn | 11:33:11 |
| 10 May 2022 |
|  Piper McCorkle (they/them or she/her) changed their display name from Piper McCorkle (she/her or they/them) to Piper McCorkle (they/them or she/her). | 05:25:49 |
| 28 Oct 2021 |
| Jamie | (cloudflare etc) | 11:33:33 |
 moritz.hedtke | In reply to @sandro:supersandro.de
Also minio the only S3 server I know of is licensed under AGPL I think ceph is also s3-compatible though I don't know the details | 11:33:44 |
 Janne Heß | In reply to @moritz.hedtke:matrix.org
I think ceph is also s3-compatible though I don't know the details yes | 11:34:42 |
| Janne Heß | Just found this (not sure if this is already known): https://www.hetzner.com/dedicated-rootserver/matrix-apple | 11:59:27 |
| Janne Heß | It's M1, not x86 | 12:00:23 |
