Sunday, 25 September 2011

How many Linux installations are there?

How hard can it be?

I'd like to know how many linux desktops/laptops, servers, phones, tablets and even routers there are.

What if each linux distro included a package that was part of a default desktop or server installation and this package submitted machine data to an online database?

A database already exists that oddly goes by this URL:

So, we have a database, we just need automated data to populate it.

We might need some rules:

1. Opt-in
2. Don't use serial numbers

Following these rules, a linux installation could send in these values:

1. CPU type
2. RAM
3. HDD size
4. linux kernel
5. Desktop, Server, Phone, Tablet, Router (based on installation)
6. Install date (of package)
7. Distro identifier

8. Number of user accounts
9. Public or private IP address (not actual IP address, just type)

A key could be generated using data 1 to 7 - things that are unlikely to change too often.

This data could be sent in monthly to either update a database entry, or to create a new one.

Old entries that don't get updated for 1 month get marked as stale.
Stale entries get deleted after 6 months.

A machine that undergoes a significant hardware or software change will automatically get a new entry and it's old entry will go stale and eventually be deleted.

At any time we will have a lower bound on active linux installations - I'd like that.