patrick

FYI, if you really think that's enough then you should check out https://feddit.org/post/2600584

The most efficient large instances cost ~$1.40 per user per year for hosting costs, and that's if you value the admin/mod costs at $0

Looking at just the hosting costs is actually a really bad indicator of total costs. The unpaid volunteer time just to run/manage the instance are likely going to be significantly more than the hosting costs if they were compensated even at minimum wage.

Each of the stacks for XXXiver.se and Bestiver.se (Mastodon + Lemmy + Static Site (+ Linkstack/Wiki for XXXiver.se premium)) are shoved into a Hetzner server at ~$13/month, and backed by R2 Object storage.

My current total hosting costs are ~$30/month to host 2xMastodon, 2xLemmy, 2xStatic Site, 1xLinkstack and 1xWiki. This is basically the minimum cost for me to host all of that on their own infra. I have approximately 0 users other than myself yet, so there's not really a useful cost/user and I can't really provide info on scaling.

Unlike most others here I'm seeing if I can make hosting into more of a job by selling the full suite of services to communities (e.g. get your own Mastodon + Lemmy + others) or by up-selling to premium accounts. I highly doubt that it will actually make any useful amount of money but I'm curious enough to try.

Hmm, I could have sworn I had code for this but I'm not able to find it. I wrote a DLX impl many years ago and used it for a few things, and I wrote several different sudoku solvers, but I don't seem to have ever used my DLX impl to solve sudoku puzzles...

What you need to do is create a row for every possible entry and location in the puzzle. So you will have a row representing every single possible entry option. 9 options x 81 total squares = 729 total rows.

The columns in your Exact Cover Matrix represent all the different constraints, where each column must be unique in the solution.

• You'll have 81 columns that represent just the location (you can only have 1 number in each of the 81 boxes).
• For every Row/Column in the Sudoku Puzzle, you will have 9 columns to represent the 9 different numbers. (e.g you can only have a single "5" in every Row of the Sudoku)
• For every 3x3 box in the Sudoku puzzle, you'll also have 9 columns for the 9 different numbers.

So your Exact Cover Matrix will need 324 columns = 81 (squares) + (9 (numbers) * 9 (rows)) + (9 (numbers) * 9 (cols)) + (9 (numbers) * 9 (boxes))

When you fill out all the rows, you'll place 1's in all the columns that that specific entry aligns with. Take the example of the row corresponding to the entry "5" in the Sudoku Puzzles top left box. That row in your Exact Cover Matrix will contain:

• A 1 in the column representing that specific box.
• A 1 in the column that represents the number 5 in the first Sudoku Row.
• A 1 in the column representing the number 5 in the first Sudoku Column.
• A 1 in the column representing the number 5 in the top left Sudoku Box.
• 0's everywhere else

To feed a specific puzzle into your solver, it kinda depends on the solver, you just need to force the output to contain those specific rows.

Your last sentence is unclear, is that actually implemented?

I am trying the approach of just making multiple affiliated services, and forcing people to have consistent account names across each. See https://bestiver.se