💾 Archived View for gemini.susa.net › discovery.gmi captured on 2021-12-05 at 23:47:19. Gemini links have been rewritten to link to archived content

View Raw

More Information

⬅️ Previous capture (2021-11-30)

-=-=-=-=-=-=-

Discovery in Geminispace

Here's the gist: "discovery should be a tier-1 discussion", and "if it can't be found, it doesn't exist".

I think Gemini offers a great opportunity for content discovery. Its simplicity makes it a relatively poor target for commercial exploitation, yet it can serve as a launchpad to good quality content, Gemini or otherwise.

This is in contrast to the web which seems to offer every incentive to game 'discovery sources', and provides every means to actually do it.

I rarely use Gemini search engines (i.e. GUS-based), because I rarely have specific questions that are likely to be answered in Geminispace. The exception is when I need to refer to notes that I've written myself, and I have my own search built into my capsule (implemented with the Swish++ indexer).

A sitemap with full-text search

More often, I'm just looking for things to pique my interest. Aggregators work really well for discovering the latest content, there are now five aggregators that I know of:

nytpu's comitium subscriptions

Antenna

gmisub

CAPCOM (shows ~30% of feeds in any month)

spacewalk

Ew0k's latest addition, Antenna, adds the ability for me to push update-requests to the aggregator, meaning that updates are brought into view within minutes. This allows gemlogs to be used for near-term announcements, for example.

For anything that's not latest-content, I like the idea of random discovery. It's hard to game 'random' without resorting to obvious spamming - links have an equal chance of being found.

There used to be a Gemini page called '<something> Roulette' (I forget the exact name) that would give a random page from Geminispace. I really liked that, but it was never updated, and eventually most of the links failed.

Therefore, to feed my gambling addiction, I created my own, which I call Lucky Bag Links (from lucky bags I used to buy in the sweet shop as a kid). These are just hacky scripts that I'm evolving into something more formal, but most stuff can now be updated on demand. I include web links in some of them, because there's still good stuff on the web - it's just hard to identify!

I welcome suggestions from anyone who would like their own interests reflected in the returned links.

Lucky Bag Links

Beyond this, I'm thinking about building a search engine from my databases of links (which are available on request, by the way). I did something along these lines as an experiment a couple of years ago using Lucene, and the results were encouraging.

Commerce Filtered Search Engine

Some background on Commerce Filtered Search

To my mind, discovery should be a tier-1 discussion. I think we should try out ideas, however simple, crazy, or ostensibly pointless, just to see what works and why. For example, word-lists generated by NLP processing of Gemini pages - just running paragraphs through spaCy can yield interesting results, and the whole of Geminispace could be processed even on low-end hardware (on my mental TODO list).

After all, if it can't be found, it doesn't exist. Okay, there are a lot of caveats in that lame soundbite, but you know what I mean.