Re: The idea of an own L4

[Marcus Brinkmann]

At Sun, 09 Oct 2005 15:17:38 +0200,
ness <address@hidden> wrote:
> The idea of an own L4 came up in irc several times. This is one thing we 
> should really think about, IMHO.

I have thought about it somewhat, but it is really the wrong question
to ask.

Here is the right question:

What operations should our kernel provide efficiently?

This is the big question, and it requires that we have a very clear
view of the system we want to implement.  I _thought_ I had a clear
view, but that was a carefully crafted illusion.

I have a new, quite different view, but it is still sketchy.  And the
new answer to the above question looks a bit different from the old
one.

In any case, once we have an answer of that question, the second
question will be: What is the most practical way to get such a kernel?
This is an interesting technical question, but it has no meaning
unless we know what type of kernel we want.

I think one of the big remaining issues is memory management and how
to separate page out policies (can be decided upon by the user) from
page out mechanism (must be part of the UTCB).  Also, kernel object
management is pretty important to me, if we want at least orthogonal
persistence.

Thanks,
Marcus