You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Currently we rank instances by TLS grade from cryptcheck, CSP grade from mozilla observatory and then the response time for a query measured by searx.space.
This does not guarantee that the best possible instance is at the top and we should think about reworking the ranking system. Since SearXNG is a fork we can completely redo this list and do for example a list with instances and not rank them at all. Also we can require more things for an instance to do, for example to not use analytics and have a constant uptime or to somehow filter instances by the quality of results.
This is just an idea, comments are welcome...
The text was updated successfully, but these errors were encountered:
IMO it is better not rank the instances / sort engines by their location makes more sense.
The ranking has absolutely no meaning in practice, by example: a bad TLS grade (from cryptocheck, what is questionable strict) weights more than a manipulated SearXNG engine with analytics or (and this is funny) a SearXNG running on Cloudflare ..
We have users who do not know what TLS or cloudflare is, they just want to have an up-to-date running instance in their neighborhood .. with the ranking we have, we give a wrong illusion of privacy to our users.
Currently we rank instances by TLS grade from cryptcheck, CSP grade from mozilla observatory and then the response time for a query measured by searx.space.
This does not guarantee that the best possible instance is at the top and we should think about reworking the ranking system. Since SearXNG is a fork we can completely redo this list and do for example a list with instances and not rank them at all. Also we can require more things for an instance to do, for example to not use analytics and have a constant uptime or to somehow filter instances by the quality of results.
This is just an idea, comments are welcome...
The text was updated successfully, but these errors were encountered: