meiqimichelle
commented
5 years ago There was a poster session related to this at IPFS Camp. I believe @lanzafame and a few others were involved -- anyone have linx to the output?
Open andrew opened 5 years ago
meiqimichelle
commented
5 years ago There was a poster session related to this at IPFS Camp. I believe @lanzafame and a few others were involved -- anyone have linx to the output?
As people start to use IPFS to pull down their packages, the traditional download counts that centralized registries provide will become less useful for measuring the popularity of packages.
One potential alternative is the number of peers in the swarm (not sure if that's the right IPFS terminology?) that are seeding (borrowing that term from bittorrent) a given version of a package, that number could also be tallied for all the available versions of a package, or perhaps just the most popular or latest.
Another alternative would be to introduce opt-in reporting of installations to some package analytics service (homebrew, which doesn't have a regular hosted registry, uses google analytics for example), perhaps in a decentralized way, although that's much easier to game than looking at currently connected peers, but doesn't require constantly asking the DHT for stats.
There's also potential privacy improvements in this metric collection change that could be listed as a a feature of using IPFS for package management, most centralized registries (npmjs.org, rubygems.org, pypi.org etc) don't allow you to opt-out of download metric collection.