Hacker News
points by abetusk 18 hours ago | 0 comments
Yeah, well, one reason the copyright information would be useful is because it's hard to come by in a standard way.

Soundcloud occasionally has copyright information tied to an individual track. Bandcamp has a structured field that lists the copyright.

A long time ago, I scraped Bandcamp metadata and put it in a SQLlite file [0]. From that you can filter on bands by copyright (among other things).

Maybe a stripped down LLM could handle the free form text and figure out licensing issues. It looks like your list is pretty reasonably sized, so that might not be such a big job.

[0] https://archive.org/details/thisonetime

[–] ferrarosimone 16 hours ago
Oh man you're opening a can of worms citing Bandcamp! It was my first pick for obvious reasons (the catalogue metadata, the fact that artists get paid more fairly, that previews may have been longer than 30secs...) but Bandcamp doesn't allow you to scrape anything anymore. Not only they block you, they formally forbid it. So I sent a "formal" request to get permission with no success. Doing that kind of 'exposed' me so I don't think I can't use their data, even someone else's scrape.

On the LLM idea: I thought the same to be honest, but the problem on Discogs is less fetching and structuring the free text and more that the info is rarely there. I launched an LLM to check and Creative Commons is there on about 1 release in 10,000, so there isn't much to rely on.

[–] abetusk 13 hours ago
Bandcamp discontinued their API but you can still scrape their site, so long as you're polite about it. They rate limit pretty severely but it is possible. The individual pages now even have JSON data that has all relevant metadata.

All major search engines index Bandcamp, so I'm not sure they really can justify shutting down everyone else.

From what I remember, about 1% to 5% of artists or songs on bandcamp had libre/free (CC0, CC-BY, CC-BY-SA) terms. I would guess that the most likely reason it's so high compared with your number is that Bandcamp has a dropdown option, essentially, so they make it easy for bands to provide alternate licensing terms. Any libre/free licensing is going to be pretty niche and the people that care about such things is also niche, so I hear you that it's a lot of work for a very small payoff.

Consider the data set size, I was, at one point, considering just downloading all libre/free data, including songs, to avoid bit rot. Maybe I still will.

Doing a quick search with that data set I linked to, I see a couple dozen dub techno bands that are libre/free licensed.

Anyway, keep up the good work. I personally enjoy dub techno very much. I don't think I've ever met anyone offline who also likes it so it's nice to see other people who are passionate about it.