> From: Don Guinn <[EMAIL PROTECTED]>
>
> > > Swapping is horrible in Windows! . . .
>
> fragmented. But the problem I am concerned about is when real memory is
> overloaded. I liked a system which wrote pages which had not been referenced
> for a while to the swap file at low priority. Then if more real memory is
> needed all that is necessary is to mark the virtual page as invalid;
> however, if the application references the virtual page before more real
> memory is needed, it is not necessary to read the virtual page back in. I
> don't think that Windows does this. It seems to wait until it needs more
> real memory before it writes pages to the swap file.
It isn't easy to judge and I don't think that Chrome tries to tackle
Windows-specific issues of underlying OS mapping between virtual
and physical memory.
Chrome addresses the swapping problem on application-specific level by
placing each tab in its own process:
(1) logical address space fragmentation for larger processes, so it
can just run out of addressable 32-bit space and (2) the physical burden
of keeping the chunks of single process memory together and related swapping.
BTW, IE has an option of launching independent instances for
different invocations, and FireFox does not seem to have this. So if
one instance crashes, in IE others stay afloat and in FireFox everything
is lost. The same applies to address spaces, their fragmentation etc.
Chrome takes IE approach to the level of tabs, so hopefully when one
tab freezes or crashes the others remain responsive.
> > > But the real problem that Chrome still does not address is that no one is
> > > really addressing the problem of software installations. It looks like
> > > Google touches on it but doesn't solve it. When a new software package is
> > > installed a cryptic message pops up in a box which the documentation
> > tells
> > > you to ignore and reply "allow". Until an operating system is defined
> > which
> > > does not allow any modifications to it even for device drivers, which at
> > the
> > > level it can do anything added software will always be able to add its
> > hooks
> > > and do who knows what.
> >
> > I would think that Installing software on the underlying OS is an
> > underlying OS issue.
>
>
> I agree. What bothered me was that the cartoon strip implied that Chrome
> addresses this issue. It can't.
----------------------------------------------------------------------
For information about J forums see http://www.jsoftware.com/forums.htm
