@orinayooye@3eyedpixel True, but we were more diligent about quality software pre-AI. We used to have a special relationship with the code. Mechanical sympathy, efficient use of resources were the things we used to care about. No one reviews PRs anymore. AI does everything.
I wrote an ePUB parser and I put a hard cap on the overall size with a static allocator, with a 2GB budget. I also put a hard cap per entry. Each inflate per entry can only occupy up to 64MB max. I think these are realistic for commercial ePUB files.
Go changed the implementation behind https://t.co/lmv8w4mSLU to a hash trie.
The API is still the same, but this is a good time to revisit how https://t.co/lmv8w4mSLU works, what happens internally, and when to use it instead of a regular map protected by a mutex.
Detailed walkthrough here: https://t.co/2KKhBMxWtE