These numbers are kind of weird and not what I expected.
They report 'git init' as over 64k. But the 'git init' size on my Git (2.54.0) is 27259, less than half that size. If you omit the unused sample hooks files it's 471 bytes, which is around what I expected.
Then they show after committing two new files the repo growing by 8kb each time, but I repeated the steps they showed and mine went 27259, 27994, 28554, which is a growth of 735 bytes then 560 bytes, which is also what I expected.
I wonder if they are accidentally measuring file system block usage instead of file byte size? That was I first suspected but the -b flag they pass to du is supposed to count bytes. (I used gdu -b in the above because Mac du doesn't have a b flag, and its numbers are consistent with the total from find -type f | xargs wc .)
I wonder if they have some hooks or something that are attaching extra metadata to their commits, maybe?
evmar | 7 hours ago
These numbers are kind of weird and not what I expected.
They report 'git init' as over 64k. But the 'git init' size on my Git (2.54.0) is 27259, less than half that size. If you omit the unused sample hooks files it's 471 bytes, which is around what I expected.
Then they show after committing two new files the repo growing by 8kb each time, but I repeated the steps they showed and mine went 27259, 27994, 28554, which is a growth of 735 bytes then 560 bytes, which is also what I expected.
I wonder if they are accidentally measuring file system block usage instead of file byte size? That was I first suspected but the -b flag they pass to du is supposed to count bytes. (I used
gdu -bin the above because Mac du doesn't have abflag, and its numbers are consistent with the total fromfind -type f | xargs wc.)I wonder if they have some hooks or something that are attaching extra metadata to their commits, maybe?
rjzak | 6 hours ago
Nice to see a shout-out to Julia Evans' work https://jvns.ca/blog/2026/01/08/a-data-model-for-git/