Re: [dev] Lessons Learned from 30 Years of MINIX

From: Anselm R Garbe <garbeam_AT_gmail.com>
Date: Sun, 6 Mar 2016 15:55:45 +0100

On 6 March 2016 at 15:39, <stanio_AT_cs.tu-berlin.de> wrote:
> * Anselm R Garbe 2016-03-06 14:20
>> On 6 March 2016 at 13:47, robin <robin.a.t.pedersen_AT_gmail.com> wrote:
>> > Microkernels defenitely resonates with the unix philosophy "do one thing and
>> > do it well".
>> > Having everything, except FUSE and such, in the kernel is doing more than
>> > one thing.
>>
>> IMHO the architectural decision for a monolith (==static linkage)
>> often reduces the management bloat of dynamic loading or similar
>> "extension" approaches that you end up with in a microkernel
>> architecture.
>
> you have dynamic vs static linkage problematic either way. that's
> orthogonal.
>
> microkernel is for me more about privileges and critical sections.
> generally speaking, it definitely sucks less to have by design less
> things that can badly go wrong.

I agree, though this is not stricly a problem with monolithic approaches either.
In the end you have to trust the process governor to limit the
privileges of a process (or thread) to its specification/requirement.
If the governor is only one SPOF (monolith) or if there are multiple
forks of such a SPOF (microkernel) makes no big difference in the end.
If the SPOF has a flaw, the flaw is everywhere, regardless the process
model.

-Anselm
Received on Sun Mar 06 2016 - 15:55:45 CET