Re: Re: Debian BSD.. cool idea
On Sun, Jan 30, 2000 at 10:27:34PM -0500, Raul Miller wrote:
> > This is presuming that the issues can't be dealt with in libc [or in
> > the linux compat library.]
On Mon, Jan 31, 2000 at 04:16:37PM -0500, Dan Papasian wrote:
> Every issue can be delt with, in some way. If you'd like to fix the
> relationship between the kernel so that mismatched worlds and kernels
> aren't always terrible, feel free.
At the moment, none of my BSD boxes are development machines. If that
changes, I'll take a look at this issue.
> > Anyways, given this supposedly wonderful support for linux binaries,
> > perhaps it could be built into something stable enough to support
> > multiple bsd kernel versions?
> What is "it"? I'm not too sure what you are referring to.
"it" refers to the previous noun clause. In this case that would be this
"supposedly wonderful support for linux binaries".
> As for speed, that depends on the application. AFAIK Linux still lacks
> any form of preemptive swapping in the VM subsystem. Result: when your
> demand for memory exceeds the physical RAM, your inactive pages must
> swap out as soon as the newer process requests them, with a result of
> heavy I/O when you first start swapping.
I believe this is false as of Linux 2.2. I seem to recall Stephen
Tweedie spending some significant time tuning this aspect of the system
prior to the release of 2.2.
The basic strategy is to swap aged pages out, but leave enough information
behind such that if that page is hit before the underlying memory is
reused it can be trivially made available again.
[Disclaimer: There *are* problems with Linux -- there are significant
subsystems which need to be re-engineered and rewritten. Disclamer
on disclaimer: However, I see those as subsystem issues rather than
architectural issues. Disclaimer on disclaimer on disclaimer: The
closest thing to an architectural issue that's been pointed out to
me is that Linux doesn't follow the BSD architecture. Disclaimer on
disclaimer on disclaimer on disclaimer: but that's simply an observation
of a difference, not a fundamental flaw.]