Web lists-archives.com

Re: Installing Stretch/Testing with absolute minimal bandwith useage




On Thu 15 Jun 2017 at 06:26:55 (-0500), Richard Owlett wrote:
> On 06/14/2017 12:56 PM, David Wright wrote:
> >On Wed 14 Jun 2017 at 08:50:18 (-0500), Richard Owlett wrote:
> >>On 06/13/2017 01:10 PM, Brad Rogers wrote:
> >>>On Tue, 13 Jun 2017 12:51:10 -0500
> >>>Richard Owlett <rowlett@xxxxxxxxxxx> wrote:
> >>>
> >>>Hello Richard,
> >>>
> >>>>netinst iso. I had never experimented with just how much could
> >>>>be done with netinst without any internet connectivity.
> >>>
> >>>Probably not a lot.  See; https://www.debian.org/CD/netinst/  Internet
> >>>connectivity is expected to exist.
> >
> >Well, it sort of depends where you draw the line between internet
> >and Internet connectivity. Fifteen years ago I played about with
> >a private partial mirror where the source of the debs was via
> >commuternet, ie Zip/Jaz disks transported from my Janet-connected
> >office. The "internet" was a cat5 cable.
> >
> >>I'm wishing I had discovered the netinst iso years ago.
> >>*NOTE BENE* Up until about a year ago I had only dial-up. The only
> >>practical installation path that met my needs was purchasing
> >>complete DVD sets. I spent many hours trying to coerce Debian into
> >>doing a minimalist install.
> >
> >What parameter are you actually minimising?
> 
> For the purpose of this thread, the downloaded byte count. Period.
> 
> However understanding netinst will influence how I approach doing a
> "minimal" install. [I use quotation marks to emphasize that
> "minimal" can mean very different things to different people.]

Then I'm guessing that DVD vs netinst is not really the factor,
but finding out how to change the package selection made by the
debian-installer itself; not something I've looked at beyond the
obvious user interface offered by expert install.

> >>If I had succeeded, I suspect the result
> >>would have been close to netinst WITHOUT internet connectivity. But
> >>I've learned a lot.
> >
> >Does it make any difference to the resulting package list whether you
> >install using netinst or a stack of DVDs?
> 
> I doubt it. If there were significant differences it would indicate
> a problem.
> 
> >I had assumed that you could
> >deselect all the options in the "Software selection" screen whichever
> >way you installed, and end up with the same thing, a system with
> >~248 packages installed.
> 
> Are all 248 packages "necessary"? [Note usage of quotation marks ;]

Probably not. Take a look at the priorities of those packages.
Here there are 14 extras, 96 optional and 53 important.
Of course, you have to study what those terms mean. os-prober and
wireless-tools are extra. Amongst the optionals are initramfs-tools,
grub, various firmware, more wireless stuff. Don't forget the kernel
image too.

> A mechanics question, "How to count the number of installed packages?"
> I'd like to be able to reproducibly count installed packages.

Various ways, but I only know how to use dpkg itself (as others here
suggest) introspectively, so I don't use that; the system you're
studying might be sitting on another partition.

To see the packages on the system, the number of entries in
/var/lib/dpkg/*.list is useful; /target/var/lib/dpkg/*.list
gives you that *during* installation, of course, as /target
is where the new system is constructed.

grep '^Status: install ok installed' /var/lib/dpkg/status | wc -l
or, for Greg,
grep '^Status: .* ok installed' /var/lib/dpkg/status | wc -l
excludes what's not yet installed if that's important.

grep -B3 -A2 '^Priority: extra' /var/lib/dpkg/status | less
gives you the names and approximate sizes of the extra packages.

I'm sure you can come up with better patterns.

And you'll probably use aptitude's why/why not feature a lot to
see how a package's gross size is influenced by the dependencies
it pulls in.

Cheers,
David.