Post AXydNStew0FDXkqZN2 by SamJSharpe@mastodon.me.uk
(DIR) More posts by SamJSharpe@mastodon.me.uk
(DIR) Post #AXyXUwD1kbcOkx5PFo by simon@fedi.simonwillison.net
2023-07-23T04:58:02Z
0 likes, 0 repeats
Just stumbled across https://github.com/pypi-diff via a GitHub code search for a keyword in one of my projects"Daily generated diffs for pypi package releases"How could that possibly fit in a repo on GitHub? They have a soft limit of around 1-5GB per repo...Turns out pypi-diff creates a brand new repo for each day to store the diffs from that day!Looks like it's been running since April - I wonder how long it can keep going like this https://abbbi.github.io/pypidiff/
(DIR) Post #AXybalXiihoxEIndqq by glyph@mastodon.social
2023-07-23T05:44:15Z
0 likes, 0 repeats
@simon hats off to the author here for gaming the system in this incredibly bold way. I wonder at what level Github considers this sort of thing "abuse" though?
(DIR) Post #AXydNStew0FDXkqZN2 by SamJSharpe@mastodon.me.uk
2023-07-23T06:04:16Z
0 likes, 0 repeats
@simon in theory "forever" because GitHub made the number of public repos an account can own "unlimited".In practise, I bet it's not unlimited, but even if it's *only* 2^16, that's still 180 years 😂
(DIR) Post #AXydiBckLjEy3Pohua by simon@fedi.simonwillison.net
2023-07-23T06:07:59Z
0 likes, 0 repeats
@SamJSharpe Hah, I'm glad it's effectively unlimited because I'm up to 664 public and 92 private just on my personal account already! https://github.com/simonw
(DIR) Post #AXyektWmBgKCkhsabw by aburka@hachyderm.io
2023-07-23T06:08:45Z
0 likes, 0 repeats
@glyph @simon It's clearly abuse and I imagine there will suddenly be a limit on the number of repositories one can create, or a combined size limit, so I wouldn't really tip my own hat to someone trying to ruin things for everyone but whatever
(DIR) Post #AXyekuTGgD73g7FImW by simon@fedi.simonwillison.net
2023-07-23T06:19:47Z
0 likes, 0 repeats
@aburka @glyph I've seen GitHub take a pretty diplomatic approach to this kind of thing - I know a few people who've had polite emails from their support team asking them to stop doing things, as opposed to sudden bans or rule changes
(DIR) Post #AXyekwcufK14Mbm25o by aburka@hachyderm.io
2023-07-23T06:11:41Z
0 likes, 0 repeats
@glyph @simon What's the point of this anyway? Is this saving some kind of ephemeral data? Or could I trivially generate the diff for any package I actually care about?
(DIR) Post #AXypaC0yDCLwOo6Nyi by lewiscowles1986@phpc.social
2023-07-23T08:20:19Z
0 likes, 0 repeats
@simon is this "cool, but abusive" tech?It very definitely feels abusive. And what would the use-case be?As a curious person, I need to place significant drain on community resources, so that I can keep diffs I couldn't possibly read.Monitoring individual packages makes a lot more sense; but still seems odd.And what happens if multiple releases go out in a day?
(DIR) Post #AXyyCSa1ry68PhG6O8 by bocytko@hachyderm.io
2023-07-23T09:56:57Z
0 likes, 0 repeats
@simon diffs linked to release notes or concise feature descriptions can be great training data for LLMs focused on coding/refactoring.
(DIR) Post #AXzR5nhFrEZKY5obYm by simon@fedi.simonwillison.net
2023-07-23T15:21:09Z
0 likes, 0 repeats
@lewiscowles1986 I believe it's scanning for diffs every ten minutes so multiple releases per day would still be captured - the repo thing is really about sharding all of those diffs in a way that avoids hitting the repository size limit
(DIR) Post #AXzvk5QyKKwDliKDQ0 by lewiscowles1986@phpc.social
2023-07-23T21:04:34Z
0 likes, 0 repeats
@simon that sounds very hostile to the network hosting, and the target network to me.
(DIR) Post #AY0Oder97wsuGvr34K by dolmen@mamot.fr
2023-07-24T02:28:16Z
0 likes, 0 repeats
@simon @SamJSharpe I'm glad that creating organizations is also free because that's how I'm trying to manage my repos. I have so many forks.GitHub should improve a user/org home page to split activity on forks vs source repos though.
(DIR) Post #AY3US06rbHrgsWa7uq by objectObject@hachyderm.io
2023-07-25T14:17:22Z
0 likes, 0 repeats
@simon I was hoping it was a tool to show diffs between versions of a single package. I often do that by looking at tags on the git repo, but those aren't always accurate or complete.