Re: Reviewing Hurd-on-L4 (and considering its future?)

[Rick Hodgin]

Forwarding my two replies sent directly to Paul and RMS to the list.

-- 

Rick C. Hodgin

----- Forwarded Message -----

From: Rick Hodgin <address@hidden>

To: Paul Boddie <address@hidden>

Cc: Richard Stallman <address@hidden>

Sent: Wednesday, May 23, 2018, 2:03:23 PM EDT

Subject: Re: Reviewing Hurd-on-L4 (and considering its future?)

Paul,

Here is the part of the original email chain (we had several back-and-forth).

I had written him asking for his advice on replacing the HURD kernel.  I

can't find my original sent emails ... I may have purged or archived them

at some point, only these replies:

-----[ Begin original email ]-----

-------- Original Message --------
From: Richard Stallman <address@hidden>
Sent: Tue, Jun 19, 2012 06:21 PM
To: Rick C. Hodgin <address@hidden>
CC: 
Subject: Re: Replacement for the HURD

    With either Linux or the HURD (as the HURD stands today) we cannot ever 
    get there.  We cannot ever get to where we have a truly free operating 
    system (kernel and up), even if we have the outward appearance and 
    illusion of it.

I don't think that follows.  Both Linux and Hurd _are_ free kernels.
The problem is that we lack free drivers and firmware for many peripherals.

Replacing the free code of Linux with some other free code won't get
us the drivers or the firmware.  That requires writing drivers and firmware.

I simply don't see how replacing the free code of Linux with other free code
would make progress along that dimension.

    I realized at that time how easily I could write a competitor to 
    Blender, or a 3D CAD system.  I even looked into it a little bit, but 
    saw how many CAD programs there were out there and thought I could not 
    write one sufficiently better than what's available to make it worth the 
    amount of time it would require.

The existing free CAD programs have a license incompatibility with GPLv3,
so it would be very useful to write a replacement.  However, one of them
is trying to eliminate its license incompatibility, and that helping
that project do so would be very useful.

-- 
Dr Richard Stallman
President, Free Software Foundation
51 Franklin St
Boston MA 02110
USA
www.fsf.org  www.gnu.org
Skype: No way! That's nonfree (freedom-denying) software.
  Use Ekiga or an ordinary phone call

-----[ End original email ]-----

-- 

Rick C. Hodgin

On Wednesday, May 23, 2018, 1:54:45 PM EDT, Rick Hodgin <address@hidden> wrote:

Paul,

I contacted Richard Stallman in 2012 and asked him about continuing

development on the HURD kernel.  He told me that it wasn't really need-

ed anymore because we have Linux.  I tried to convince him that GNU

needed its own kernel and we shouldn't rely upon Linux, but he told

me that instead of spending time working on a GNU kernel, I should

move to work on making a compatible Adobe reader, and other such

practical applications.

I don't think there's any interest from Stallman down on working on a

GNU kernel.  I think that's a mistake, by the way, but to my knowledge

that's where we're at.

I've cc'd Stallman on this email so he can verify his position.

-- 

Rick C. Hodgin

On Wednesday, May 23, 2018, 1:02:55 PM EDT, Paul Boddie <address@hidden> wrote:

Hello,

The day before yesterday, I found myself editing the L4 page on the Debian

Wiki:

https://wiki.debian.org/L4

It occurred to me that in the context of the Hurd, very few updates about L4

implementations have been propagated via various channels for probably a

decade or more. This meant that the above page was outdated as well as

inaccurate or incorrect on some matters (and it wasn't exactly a long page,

either).

Reviewing the archives for the l4-hurd list, I see that very little discussion

has taken place since 2010 or so, where Viengoos was once again mentioned. And

the background to all this is, of course, provided on the Hurd site:

https://www.gnu.org/software/hurd/history/port_to_another_microkernel.html

What I would like to know is whether more modern L4 implementations might be

reasonable candidates for attempting this kind of work again. It seems to me

that L4 implementations of the early 2000s were considered inappropriate or

insufficient for the design of "Hurd-NG", leading to things like Viengoos.

But then again, L4 implementations have since been widely deployed, albeit

perhaps in very specific roles. Meanwhile, environments like L4Re and Genode

must surely go some of the way towards providing the basis of something like

the Hurd.

The hurd-l4 work that was previously released got as far as running the banner

program, apparently, but surely things like L4Re and Genode go quite a bit

further than that. Obviously, they are merely the foundations of any potential

Hurd-like system, but it would not be unfair to claim that anyone wanting to

implement such a system would start off with a lot more code already written

than was the case fifteen or so years ago.

So, has anyone seriously considered a fresh approach to this endeavour using

more recent L4 implementations and environments? I largely ask because I have

been developing programs to run on certain niche hardware devices using L4Re

and the Fiasco.OC microkernel, and although some aspects of this are

frustrating, it is surprising how much supporting code there is already. I can

envisage building components on top that support a more general computing

environment.

Does anyone have any constructive perspectives on this?

Paul