Regular expressions considered harmful.
https://github.com/COSI-Lab/Mirror/blob/046b562660d58b747073618f947ff0b8abb67ca0/nginx.go#L119-L120
Splitting on the " characters is significantly more efficient, somewhere between 10x-100x so. This kind of performance improvement wouldn't be meaningless either. Somewhere between 500 MB and 1500 MB of data is passing through this parser daily.
This would make a good first issue 😉. You can easily compare the output of the previous version to a new version on gigabytes of data.