[Microkernel-devroom] FOSDEM 2013 - Microkernels / Component-based OSes devroom

Stefan Kalkowski stefan.kalkowski at genode-labs.com
Mon Dec 3 13:04:39 CET 2012


Hi folks,

sorry for my very long absence.

On 10/30/2012 08:13 PM, Joanna Rutkowska wrote:
> On 10/30/12 11:16, Martin Decky wrote:
>> Hi Joanna, hi folks,
>> 
>>> So, perhaps during this meeting there could be a session which
>>> would aim to create such a big picture diagram or something.

I have to admit, to me this appears to be a bit contradictory to the
idea of FOSDEM. I would like to focus on finding common problems,
discussing different approaches to target them, learning from the
experiences of other projects, identifying synergies.

In my eyes, developing a taxonomy about: who is in competition to
whom, and what are its differences, is more about creating an overview
for potential customers of the projects, than discussing common
technical issues of it.

>> That's a good idea. Last year there was a closing panel
>> discussion on the microkernel devroom which might be a suitable
>> place to organize such a "big diagram drawing" this time. Or
>> would you prefer a separate session for it?
>> 
>> Anyway, we should perhaps brainstorm the interesting criteria
>> according to we would like to classify the various systems on
>> this mailing list first, so that the "big diagram drawing" is
>> somehow structured when it happens live. Alternatively, I can
>> host a wiki page for it (on the HelenOS wiki).
>> 
>> Therefore, in relation to Joanna's suggestion, I suggest to have
>> a closing panel discussion again on the devroom. I also suggest
>> to allocate a longer time slot for the panel discussion this time
>> (at least twice as long as last time).

Although, I enjoyed this year's panel discussion, we're limited to one
day (Saturday) with regard to the room. Moreover, there will hopefully
be more active devroom members this time. At least, this is what I got
signalled before submitting the devroom application. To give most of
them the chance to present some work, I would reluctantly reduce the
time for presentations. Of course, there will be the possibility to
have informal discussion also on Sunday.

My personal proposal is: having a "demo" slot, where all projects can
have a laptop demonstration, or poster. We ourselves, or guests of the
devroom can travel from table to table discussing key aspects of the
projects. In my perception, it is a lot easier for most people to have
a conversation, or ask their questions this way than having a podium,
which always has a more formal atmosphere, whether it is intended or not.

> I like the idea with a wiki. In the first iteration people from 
> different projects that we consider here could first create some
> short stubs describing key aspects of their projects (features,
> assumptions, requirements, dependencies, etc). Then, in the
> following iterations we could try to come up with some readable
> structure for this.
> 
> E.g. I think I could write something like this about Qubes OS to
> start with:
> 
> * Category: Desktop OS (available as installable ISO) * Goal:
> reasonably secure multi-domain desktop environment with lots of 
> mechanisms to make this as seamless experience to the user as
> possible. * Main focus on: 1) secure GUI, 2) secure inter-domain
> interaction mechanisms, 3) reasonable TCB decomposition (netvm,
> usbvm, etc) 4) domain management * Depends on: Xen, Fedora/Linux
> (VMs), Xorg * App compatibility: Most Linux Apps except for such
> that require accelerated graphics * Drivers compatibility:
> Essentially all supported by Linux kernel unless they break when
> run under Xen (some GPU drivers do :/) * Status: Qubes 1 released,
> Qubes 2 Beta 1 ISO in the work (with HVM and specifically Windows
> VM support, most code in the Git already). * Non-goals: creating a
> new microkernel (rather how to best use existing ones), creating
> drivers, creating new user apps.

Regardless of my personal objections, I don't want to choke your
enthusiasm, nor overrule Joanna's proposal. Maybe, we wait for the
final number of proposals and decide then, whether to develop such a
diagram in an official time slot on Saturday, or do it informally
maybe on Sunday?

Here is the picture taking the same categories like above for Genode:

* Category: OS Framework
* Goal: framework to build OSes for special use-cases to a general
purpose OS with particular attention to security
* Main focus on: reducing the application-specific TCB, microkernel
techniques with dynamic workload
* Depends on: Linux | Fiasco.OC | NOVA | OKL4 | Pistachio | Fiasco |
Codezero
* App compatibility: POSIX compatibility layer (Noux) enables porting
of Unix applications, Linux VM enables execution for unmodified Linux
binaries
* Drivers compatibility: different classes of Linux, OSS, or iPXE
Device Drivers by corresponding compatibility layers: USB, Audio, Nic,
Gallium
* Status: Genode 12.11 (3 months release cycle)
* Non-goals: reinvent the wheel for each subproblem (don't create new
drivers, protocol stacks, or microkernels as long as usable exists)
* Target Audience: system developers, not end users (folks like Joanna
;-))

Best regards
Stefan

-- 
Stefan Kalkowski
Genode Labs

http://www.genode-labs.com/ · http://genode.org/


More information about the Microkernel-devroom mailing list