• 5 Posts
  • 21 Comments
Joined 3 years ago
cake
Cake day: December 28th, 2021

help-circle











  • It’s cool that they have that info (hadn’t noticed it before), though it only looks at the front end code, there is no way to check for any malicious code on the back end. At the very least, logging the IP addresses that use the search engine is trivial without any changes to the front end. Actually correlating search queries to IPs would probably be harder since the query isn’t in the URLs, but maybe not that much harder, I’m not sure.



  • As you’ve said, the major issue with SearX and other open source and self-hostable options is that they don’t have their own crawlers and self-hosting means giving your IP to the big tech search engines you’re pulling results from. Tor or other proxies are always an option to prevent this but they make it more likely that your search requests from Google, etc. will be blocked. For centralized search engines with better-than-average privacy policies and their own crawlers, Brave Search probably is the best option at the moment in terms of returning relevant results.

    That said, I’m eagerly awaiting for an open-source, self-hostable search engine with its own crawler. We shouldn’t be satisfied with the current lackluster options.


  • DDG censors things based on politics because Bing (where it gets its search results) does. And Google also censors results based on politics so I’m not sure how SP is a step up in that department. That being said, Google’s results tend to be more relevant to the search query than Bing’s, which is why SP’s results are often more relevant than DDG’s.




  • I agree that pointing out the problem is far easier than finding a good solution. I don’t think activity-based sorting is much better since, as you said, that just tends to promote the most outrageous content. Facebook and Twitter-like platforms suffer from that issue more than Reddit-like platforms do. In short, I don’t have a good solution and I acknowledge the benefits of the upvote/downvote system (such as outrageous and irrelevant content being filtered out by the community without the need for as much active moderation), but it is a poor tool for fostering civil, ideologically diverse communities.



  • Ultimately, Lemmy is a Reddit clone and one problem with Reddit-like platforms is the upvote/downvote system heavily promotes groupthink since dissenting opinions are downvoted into oblivion while consensus opinions are promoted. Lemmy attempts to solve this problem by being open source, self-hostable, and federated (which are all great things), but these aspects alone can’t totally solve the inherent groupthink problem, it just makes it easier for those with differing views to spin off and start their own instances, which will likely have their own groupthink.

    As politically centrist myself, I’ve basically just unsubscribed from the political communities and focus on the more tech-related ones. If an apolitical or more centrist political Lemmy instance which federated with lemmy.ml was started there’s a good chance I’d join.






  • “Trying to build socialism.” As with most professed communist societies, China is currently ruled by an elite class that suppresses the majority of the citizens. Surveillance capitalism is bad and I do try to avoid Google products to the extent reasonable, but the CCP is a much bigger threat to free speech and world peace. If you think they just want to control China (which includes Hong Kong, Taiwan, etc. in their eyes), you’re deluding yourself.