💾 Archived View for bbs.geminispace.org › s › Geminispace › 5025 captured on 2023-11-14 at 08:50:29. Gemini links have been rewritten to link to archived content

View Raw

More Information

⬅️ Previous capture (2023-11-04)

➡️ Next capture (2023-12-28)

🚧 View Differences

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

Which search service do you recommend in Gemini space?

I’m currently using gemini://geminispace.info/search/ as my default in Lagrange. At the moment - and this could just be the nature of the contents of Gemini space currently - I get a lot of results for a library flooding my results. The results contain the terms I searched for, but that domain won’t have what I want.

What are the alternatives and is there much difference anyway? Please let me know your preference.

Posted in: s/Geminispace

🚀 IO

Sep 01 · 2 months ago · 👍 Addison

13 Comments ↓

🚀 IO · Sep 01 at 17:44:

The library is library.inu.red. There are other domains dominating the results too.

I can’t imagine a good way to change search behaviour myself so unfortunately I don’t have a constructive suggestion. One idea I had was that any domain could have a limit of how many times it’s represented in search results, with a link after the last occurrence titled “more results from this capsule”.

☕️ Morgan · Sep 01 at 20:26:

I don't really use search; just bookmarks, link lists and aggregators.

🚀 skyjake · Sep 01 at 21:07:

Have you given Kennedy (by @Acidus) a try? I've found it to be quite sophisticated.

— kennedy.gemi.dev/

Check out the features here:

— kennedy.gemi.dev/docs/features.gmi

🐝 Addison · Sep 01 at 21:18:

I maintain library.inu.red; I'll update its robots.txt to discourage indexing full documents and ask the owner of geminispace.info to purge and re-index that host.

I never really use search features on here so I didn't realize how overrepresented it is, but I see what you mean.

🚀 IO · Sep 01 at 21:42:

@Addison, I didn’t mean to single you out. It’s just that I’m on mobile and I haven’t figured out how to save a draft yet. I couldn’t check other tabs on my Gemini client to copy the other URLs.

There is great content on your library. I’m not sure if de-indexing is right!

🚀 clseibold · Sep 01 at 21:49:

Geminispace.info, while it was a breakthrough for Gemini for being the first search engine, and I want to give it respect for that.... it just isn't a good search engine, imho. Try TLGS, Kennedy, or that newer search engine. They should be listed in the official Gemini's project capsule, section 2 of the FAQ.

I had my own search engine, but it's down right now. I want to bring it back up soon, so that I can make the search engine that I think is best. But that won't be for a while, unfortunately. Otherwise, TLGS and Kennedy are actually pretty good, iirc.

🚀 clseibold · Sep 01 at 21:51:

@Addison Imho, the problem is not with your capsule but with geminispace.info. De-indexing would technically help, but it kinda defeats the purpose of searching to begin with. We really should just be moving over to better search engines.

🐝 Addison · Sep 01 at 22:45:

@IO Thank you for saying so; and I don't feel singled-out, so no harm done. I don't often use search engines on gemini so this wasn't something that had crossed my mind.

In my opinion it's for the best that the gemtext documents themselves are not indexed. The Topics and Authors listings will still be indexed, so folks who want to search for a particular subject will be able to find what they need with less background noise. As a reader I would rather get sent straight to '/topic/Abc' rather than have to sift through the 200 articles that happen to mention the term I entered.

🐝 Addison · Sep 01 at 22:56:

@skyjake @Acidus Kennedy is very impressive, I'm bookmarking this. Thank you for the recommendation.

🍵 michaelnordmeyer · Sep 02 at 10:57:

Kennedy has by far the most features and therefore is able to provide the best search results. If the number of results are too many, filters can realiably weed out unwanted results.

Other search engines have only a subset of the features of Kennedy and don't lead to the same reliable results.

The only problem with Kennedy is that it crawls only manually and the last crawl has been on 2023-07-23.

I created a page with meta information about different search engines and crawlers, which are not always available on the respective search engine's pages:

— Meta Information About Geminispace Crawlers

🚀 clseibold · Sep 09 at 23:38:

Both Kennedy and TLGS are good, but I've found TLGS has better search results than Kennedy, even with Kennedy's "most features" (which is a vast overstatement, imho, and not all of these features affects search results as much as one might assume). All of the search engines have filters, some of them just don't advertise this as much as the others. TLGS has features Kennedy doesn't have, and Kennedy has a couple features TLGS doesn't have. To say "Other search engines have only a subset of the features of Kennedy" is too simplistic and overstated. None of these features matter if they don't actually and practically make your search results better.

Number of features a great search engine does not make, especially if you prioritize those over how results get ranked/scored, which is one of the most important parts of a search engine. Both TLGS and Kennedy use link-popularity-derived ranking (or link-analysis, similar to PageRank or SALSA)

Lastly, GUS, Kennedy, and TLGS are mostly designed to do what is known to be practical in searching. They aren't innovating as much with different types of ranking, for example. This is fine, but this also means that the bar is somewhat low for Gemini Search Engines, so it wouldn't be very hard to make a search engine as good as Kennedy and TLGS.

GUS (aka. geminispace.info) is the lowest bar out of all of these, so if we're just comparing our favorite search engine to GUS, then you're already getting a skewed/biased conclusion, and so it doesn't mean much.

Anyways, none of this is to hate on Gemini search engines. They are doing a pretty decent job. But we can do better. No, Kennedy is not "sophisticated". Its features are fairly basic.

🧇 Acidus · Sep 15 at 14:31:

It's always fun to play with different approaches to indexing and surfacing data. Honestly the "just play" elements of it are one of the main things I love about Gemini.

I built Kennedy for myself as a fun project, and I've just keep adding features for things I'm interested in. That's how the Delorean Time machine (like Archive.org, but for Gemini) came to be as well. Recently my interest has been around trying to surface non-text content more effectively. It's also lead me to some cool content that I wouldn't have found other wise.

I'm so pleased with all the positive feedback on Kennedy! Comments like that make my day ❤️ I'll keep working to make it better for everyone.

🧇 Acidus · Sep 15 at 14:31:

oh, and special shout out to @michaelnordmeyer who has reported some many bugs and other issues over the years.