| 16 Oct 2025 |
 raitobezarius | I said 5.4M packages data per month | 11:09:51 |
 BMG | Half way there 😉 | 11:10:11 |
| BMG | Looks about 10 millionish a month | 11:10:20 |
| BMG | I can do the month rollup | 11:10:26 |
| raitobezarius | And given that it's narinfo, packages data should be twice that amount or so | 11:10:38 |
| BMG | I can tell you the exact size | 11:11:00 |
| BMG | I have all that info | 11:11:02 |
| BMG | not just for narinfos | 11:11:09 |
| BMG | actually I made a mistake | 11:11:47 |
| BMG | That query is all objects, not just narinfos | 11:11:55 |
| raitobezarius | so I win | 11:12:03 |
| BMG | Let me do a few separate queries here, be more careful | 11:12:07 |
| raitobezarius | But basically the calculation is there's roughly 300k derivations per nixpkgs eval across the 4 architectures, there's 6 channels in parallel (some have common derivations, etc.), so it gives you ~1.8M upperbound over the 6 channels, given that the fast paced channels move every 1-2 days and the slow paced channels move — well — when they want, you can reasonably say there's some interpolation of 1.8M * 3 = ~5.4M derivations happening (and I said every month, but I should have said every 2 weeks or 3 weeks) | 11:14:39 |
 vcunat | With this kind of computation it's easy to make 10x error in the totals. | 11:18:33 |
| BMG | 
Download image.png | 11:40:32 |
| BMG | This is filtering for just narinfo | 11:40:41 |
| BMG | 
Download image.png | 11:42:12 |
| BMG | This is filtering for '%.nar.%' | 11:42:20 |
