Speeding up the debian/etch boot, a short report

[Petter Reinholdtsen]

[Marco d'Itri]
>>   S02module-init-tools
>>   S02modutils
> These are supposed to be run with a rw root (I do not mind removing the
> depmod call on principle, but I fear some people depend on it).

Ah, right.  I'll add checkroot as a dependency for these then.  Thank
you for the heads up.

>> Next, I realise that ntpdate and portmap do not really need to run in
>> rcS.d/.  Moving them both to rc2.d and running insserv -v again, it is
> portmap does, if /usr or /var are NFS-mounted.

Why?  NFS mounting should work without portmap enabled on the local
machine, right?  nfs-common is started later with the stat and lock
daemons, so these do not need portmap.

> And if somebody cares enough to run ntpdate on boot then probably
> they want to run it as early as possible (and it has to be slow, by
> design).

Perhaps.  I didn't quite make up my mind here.  On one hand I would
not really want ntpdate to run before I get the shell in single user
mode (when I need singleuser, I do not want to wait any more than I
have to), and on the other it is nice to have a correct clock as soon
as possible.

> Indeed, hotplug does.

Hotplug didn't load any of these modules on the machines in question,
even though it does have an IDE CD-ROM and a PS/2 mouse.  Should it?

> Another possible optimization is to control the fixed network
> interfaces with hotplug too, but this may require some bits of
> infrastructure which are not yet in the package.

Do you have more info on this one?  I do not understand your
suggestion.

> Interesting work, but of limited use: modern systems do/will not work
> this way anymore.

Your initial comment almost made me disregard your entire message.
The stable release of Debian work this way, so I find the testing
results quite relevant and not really of limited use.  Your comment
tells me that you consider Debian/sarge not to be a modern system, and
to put it simple, I do not agree.  I'm not sure if this was the
intention, but thought it best to let you know what effect your
statements had on this reader.

Oh, well.  I decided to keep reading, and ignore the most extreme
comments and focus on the ones I find productive.