Boot speedup with readahead

[Ralf Corsepius]

On Mon, 2008-09-15 at 00:48 -0500, Callum Lerwick wrote:
> On Thu, 2008-09-11 at 12:33 +0200, Ralf Corsepius wrote:
> > I would turn this argument around: rpm missed its opportunities "to do
> > various things" forcing people to circumvent rpm's limitations by
> > ruck-sacking rpm with add-ons such as yum, yast/ycl etc.
> 
> Oh come on. How is layered, loosely coupled, encapsulated separation of
> concerns NOT good design?
The point is: Any actions on rpm must be performed through rpm (rsp.
corresponding library calls= only.

In other words: rpm -U|-i|-e must work. High end frontends such as
yum/smart/apt/yast must not performing additional hacks interfering into
the system (e.g. prelinking) or the rpmdata bases.

> The diversity of RPM front ends is a sign we're doing things right.
I disagree. Provided the history of rpm, I take the diversity of
highlevel RPM front ends, as a sign of diverging interest and of
influence of vendor hegemony.

> Do we want to drag all of Yum's deps (python, and so on) down in to RPM
> itself?
C.f. above. 

>  How about C++?
Implementation detail.

> RPM should be kept as small and simple as possible.
It should perform what is necessary to make rpm -U|-i|-e etc.
deterministically working.

Ralf