Caching dependency archives #31
Description
Another significant performance gain, can be had by caching the tgz files locally after first fetch and reusing that on previous fetches.
This is one of the main causes helm dependency build/update are slow when running them on a big helm repository.
It's 2-5s per chart, chick at a decent repo of 20 chart, is already 40-100s!
Now think about a repository the size of k8s-at-home and we're talking 2502-5s= 500-1250s(!)
or even worse: truecharts: 7502-5s
It looks like the Helm team decided to just drop their own use repository, partly so they don't have to optimise their hugely inefficient toolkit.