Follow

How many times a day you use a search engine (please RT).

Based on how search engines work (keep the index fresh, store the data to be queried) a pay-per-use approach is not sustainable

How much are you willing to pay for subscription for a search engine per month per person ?

(see the previous pool)

Given a search engine that is privacy respecting that you can self-host on your own hardware.

How much are you ready to spend on a home server upfront that will last 3 to 5 years and can support your whole family and friends:

I should have started with this:

I am working on a search engine . Unlike a todoapp, it requires more hardware, having an idea of how much people are willing to spend upfront or by subscription renting a server will help me figure what kind of hardware I need to acquire to be able to test the algorithms.

Thanks for you cooperation! (and please boost!)

@amirouche
I think your low numbers are too high, both for hardware and subscription.
I pay less than 10 bucks for mobile Internet. Yet, somehow, the search engines and all the ad-loaded pages (plus all the hundreds of layers of ad maf... industry) make a living from eating my bandwidth.
It should be 1€. 3 max.

@danielst

Some people are paying for those ads, hence the cost to run those search engines.

@amirouche Whoa, even 10€ seems like a lot of money to pay per month per person for a search engine. 5€ might be more reasonable...
@amirouche Pay-per-search might also be reasonable for those who don't do enough searching to justify a monthly subscription. Something like 0.05€ per search would give someone 100 searches before a monthly 5€ subscription becomes worth it.

@jbauer

Thanks to chime in!

Pay-per-search does not work as the only revenue to cover infrastructure because you need to pay for hosting / renting anyway so that it is available when someone needs it.

@jbauer

You made me remember that the server sits idle most of the day: if they are 5 persons doing 100 queries per day where each query takes less than 1 second, the server will be fully busy, all cpu core maxed out, 500 seconds per day that means there is still 85900 seconds available (~23.86 hours) on all CPU cores...

What if the server provides more "services" or "applications" like ? Would 10€ per month per users makes sense?

ref: nextcloud.com/pricing/

@amirouche In that sense you'd be providing a whole cloud service to your users. You'd be creating a subscription-based Google-like ecosystem which is a different product than simply providing a search engine to your users. It might be worth 10€+ to users who want a complete managed Google alternative. However, if all somebody wants is a good search engine, they'd get a whole cloud service offering that they probably don't want or care about so they probably won't bother with it.

Maybe a tiered subscription scheme would be useful? Something like: people can pay 5€ for search and maybe 5€ more for cloud services and 5€ more for extra storage space or something.

@jbauer

Having too much features is is not good nor easy.

We are gathering on irc at freenode the channel if you want to further discuss the matter.

@jbauer

The features only of the search engine part: etherpad.wikimedia.org/p/babel

At that does not include taking into account user feedback to be able to program the crawl from the search e.g. via a feed reader.

Also it does not mention interop between search engines which we dive in etherpad.wikimedia.org/p/peace

@amirouche I'd prefer a worker system where people contribute their bandwidth and crawling capabilities to have access
the actual amount isn't much once you get > 10 users, and it becomes a "transparent cost"

@toast

Thanks to chime in!

A worker system like you describe is much more complicated to develop. Also, in such a system people need to trust each other that they will not game the crawl results.

Bandwidth and crawling is necessary to seed the index and then keep it fresh. The initial index can be shared (like Common Crawl Search Engine does, but with less spam). Keeping the index fresh for a family is doable on a regular fiber or DSL connection, it does not consume much.

@toast

The storage is not even costly, english wikipedia + stackoverflow does not even reach 100GB.

What is costly is doing a query under one second. For that you need lots of CPU cores / threads like AMD epyc or Threadripper.

Maybe I did not understand what you wrote?

@amirouche I did have further ideas in that regard, but obviously my theoretical search engine doesn't exist yet
as far as what I *actually wrote* in this thread, you're correct
either way, I'm likely to be the host (I have a 24-core server at home) than to pay someone to be it, so I was approaching it from that direction :blobsmile:

@toast

Now that I think more about it, exchanging queries can work.

We are gathering on irc at freenode the channel if you want to further discuss the matter.

@amirouche nah I'm good
I have enough projects to worry about now, and I'll re-evaluate what's available once I get to search engines on my long todo list
I do wish you lot luck though :blobsmile:

@amirouche By average I use it 50+ times per day. I'm a developer so searching things is an integral part of my work. :morty:

@Ghosty

Thanks to chime in!

Yes me too. By the way, I found out sometime it is faster from a workflow perspective to lookup the documentation locally for instance man pages or in the case of Python using `pydoc collections.Counter.most_common`

@Ghosty

More related to online search ux, the fact that query history is hidden somewhere and there is no way to gather all search queries under an activity tab is painful. Many times, I recall doing a search query about an activity, but can not recall the exact terms. Do you experience something similar?

@amirouche Exactly the same actually! I remember that there was a problem with my xmpp server and it took hours to find a solution online.
I recently had this issue again and knew that I looked it up before but wasn't able to recall what exactly did I ask the search engine. So it took several hours to find again :I

@amirouche I use documentation too. It is a faster way to find what you need once you are "comfy" with the programming language.
A decent mixture of docs and web searches does the trick for me. :)

Sign in to participate in the conversation
FLOSS.social

For people who care about, support, or build Free, Libre, and Open Source Software (FLOSS).