[lokedhs]

I remember installing Hurd on actual hardware back in 1996[1] and I as far as I can recall, it was quite interesting to use.

Since then, on regular intervals[2] I've attempted to install the most recent version, but I have failed every single time[3].

I'm also noting that Hurd is still only 32-bit.

Is there actually a plan to make this into a system that can actually be used? Exactly what is it about the Hurd project that makes it take so long to finish? There have been several examples of single developers creating a usable[4] operating system in the span of a couple years.

[1] Yes, 20 years ago

[2] Perhaps once every one or two years

[3] Running a preconfigured VM image doesn't count

[4] More usable than Hurd is today

[disordinary]

Basically Linux and then BSD took all the wind out of Hurds sails. A lot of that was timing, and also because those projects were much more attractive to big business who put a tremendous amount of development resources behind them.

As far as I know QNX is the only usable OS based on a microkernel.

[buserror]

Errr... OSX and iOS are also MACH based. I'd call them 'usable' ? :-)

Funny bit is, back in the 90's when Apple bough NeXT, they trashed the NeXT version of MACH and used the one that was in... mklinux instead!

I bet very few people remember that bit. I remember because I had written the framebuffer console driver for mklinux back then, and seeing my init message when booting the earlier version of OSX (<=10.1 ish, perhaps a bit later too)

So I had the privilege of printing kernel crash logs and panics on zillions of devices! I'm so proud! :-)

[disordinary]

It's classed as a hybrid though, not a pure microkernel. There's significant amounts of BSD in there.

Also, thanks for the anecdote :)

[pjmlp]

Quite a few embedded OSes are microkernels, L4 being one of them.

[disordinary]

Interestingly L4 is being used to power the secure enclave on iOS devices so it's probably the most distributed microkernel.

[lokedhs]

There was a plan to move Hurd to use L3 instead of Mach, but apparently that project was cancelled.

[flukus]

> I'm also noting that Hurd is still only 32-bit.

I wonder if they'll make it to 64-bit before we start upgrading to 128-bit computers?

[daveguy]

Probably. It's unlikely we will be upgrading to 128 bit computers any time soon. 64 bit words are enough to store a double precision floating point number and enough bits to directly address 18.4 exabytes of RAM. The jump from 32 to 64 is a huge leap and any issues we were having with a 32 bit word are solved for a much longer period than the jump from 16 to 32 bit.

2^8 => 2^16 : 256x increase

2^16 => 2^32 : 65,536x increase

2^32 => 2^64 : 4,294,967,296x increase

[flukus]

I know it doesn't seem likely, but these things have a way of creeping up on you. 25 years ago the prospect of everyone having 8 GB of ram was far fetched, computers still had 8MB and 8GB hard drives were years off. In 10 years we could have petabytes in our phones and from there exabytes aren't such a huge leap.

[daveguy]

But to put it in perspective

8MB => 8GB : 1,000x increase

8GB => 8EB : 1,000,000,000x increase

The leap to 8 exabytes is 6 orders of magnitude greater than the last 25 years.

That would be 30 Moore's doublings (applied to storage) and would take 60 years. Oh, and we would still be able to address all of the ram with a 64 bit address.

In 10 years we will be closer to 256 GB in the phone, not petabytes or even terabytes (but terabytes would be close).

[brohee]

They also still publish SHA1 sums, definitely stuck in the past.