Merge remote-tracking branch 'fp/master'

Conflicts:
	open_issues/secure_file_descriptor_handling.mdwn

1 files changed, 39 insertions, 1 deletions

diff --git a/microkernel/discussion.mdwn b/microkernel/discussion.mdwn
index a5a73e18..f5626f6c 100644
--- a/microkernel/discussion.mdwn
+++ b/microkernel/discussion.mdwn
@@ -10,7 +10,7 @@ License|/fdl]]."]]"""]]
 
 [[!tag open_issue_documentation]]
 
-IRC, freenode, #hurd, 2011-07-26:
+# IRC, freenode, #hurd, 2011-07-26
 
     < antrik> Tekk_`: regarding microkernels: the basic idea, and really the
       *only* fundamental difference, is that they isolate things in separate
@@ -22,3 +22,41 @@ IRC, freenode, #hurd, 2011-07-26:
       these are secondary effects: such choices can also be implemented in a
       monolithic architecture -- and not necessarily harder. just less obvious
       in some cases...
+
+
+# IRC, freenode, #hurd, 2013-08-28
+
+    <Spyro> ok question
+    <Spyro> what is the big advantage of microkernels over monolithic kernels
+      as you guys see it?
+    <Spyro> is it entirely for the benefit of developers or are there actaully
+      practical advantages?
+    <kilobug> Spyro: there are many advantages, at least in theory, in terms of
+      modularity, flexibility, stability, scalability, security, ... which are
+      for everyone
+    <braunr> Spyro: of course some advantages are practical
+    <braunr> for me, the main advantage is system extensibility
+    <braunr> you can replace system services at runtime
+    <braunr> and on the hurd, you can do it as an unprivileged user
+    <braunr> (the direct side effect is far increased security)
+    <braunr> kilobug: i don't see the scalability advantages though
+    <kilobug> braunr: I would say it goes in par with the modularity, like, you
+      can have a full-weight IPv4/IPv6 stack for desktop, but a minimal stack
+      for embeded
+    <braunr> i see
+    <braunr> for me, it's in par with extensibility :)
+    <braunr> i see modularity only as an implementation of extensibility
+    <braunr> or a special case of it
+    <braunr> Spyro: basically, it's supposed to bring the same advantages as
+      fuse, but even more so (because it's not limited to file systems), and
+      better (because it's normally well integrated with the core of the
+      system)
+    <teythoon> also, fuse is kind of bolted on and Linux composes really badly
+    <teythoon> e.g. it is not possible to nfs export a fuse mounted filesystem
+      on Linux
+    <braunr> bolted ?
+    <teythoon> isn't that the term? as in being attached using screws?
+    <braunr> i'm not familiar with it :p
+    <azeem> "a posteriori design"
+    <teythoon> yes
+    <braunr> ok