Bug#696154: cloud.debian.org: Please install 'less' by default on official Debian AMIs.
On 27 December 2012 11:28, Charles Plessy <firstname.lastname@example.org> wrote:
> Le Tue, Dec 18, 2012 at 03:14:02PM +0100, Stefano Zacchiroli a écrit :
>> On Mon, Dec 17, 2012 at 02:11:52PM -0800, Clint Byrum wrote:
>> > How is it a slippery slope if it is driven by data?
>> > Seriously, figure out a way to ask users what they want. popcon isn't
>> > going to be all that useful here becuase of the wild diversity of systems
>> > that exist in popcon. But you can certainly just ask users to list any
>> > optional packages that they'd like to see on images. Or have a subset
>> > of popcon just for cloud images.
>> Data is always good to have so, sure, let's find out ways to do that.
>> But I urge to figure out how to gather data that distinguish the wishes
>> that are cloud-specific wrt the others.
>> For everything that is not cloud specific, I think we should strive to
>> make the corresponding improvements where they belong, i.e. in Debian
>> default installation choices. And I'm sure there is room for
>> improvements there, because there is always room for improvement :-)
>> Here, I think we should be mostly concerned for cloud-specific needs
>> and, sure enough, we should add them to the pre-built images we offer.
> Hi all,
> I think that the Debian defaults should be based on common practice. In that
> sense, I think that we should first work out a package list that suits our
> needs, and only after, if we can demonstrate that it is of general interest,
> propose that it may be reflected on Debian's standards.
> It would be tempting to use package priorities, with "important" representing
> the "bare minimum" discussed earlier, and "standard" representing the images
> that are ready to use for some simple tasks. However, this would mean downgrading
> the priority of exim4 and raising the priority of openssh. I do not volunteer to
> lead this discussion...
> I think that this rules out bothering email@example.com until we
> have a good record of providing images that are used broadly, except perhaps to
> propose a new "tasksel" task (or more if relevant).
> We therefore need a good definition of what is minimal, in terms of packages
> and in terms of image size. For instance, on the Amazon cloud, the size of
> instances is defined in gigabytes, and our images are currently configured to
> use 8 Gb volumes by default.
> For the cloud-specific part, the defintion of what is minimal also needs some
> arguments, that can for instance justify why we ship systems with ssh by
> default and not other packages, as it is equally easy to install them with
> user metadata.
> Lastly, there are packages like "less", or "psmisc" (/usr/bin/killall), that
> have a neglectable footprint in terms of cost and security. I understand the
> argument of slippery slope, but if we consider the 8 Gb images discussed above,
> there is enough space to install some of them. If we all agree that the
> contents of the images is not set on stone (that is, we can remove "less" when
> it proves to be deleterious to some users), why not satisfying our current
> users (including myself), instead of focusing on the leanest solution, that I
> think is likely to attract less users.
> Charles Plessy
> Tsurumi, Kanagawa, Japan
> To UNSUBSCRIBE, email to firstname.lastname@example.org
> with a subject of "unsubscribe". Trouble? Contact email@example.com
> Archive: [🔎] 20121227102816.GA12377@falafel.plessy.net">http://lists.debian.org/[🔎] 20121227102816.GA12377@falafel.plessy.net
Up until this instance the counter arguments had me swayed, but you do
make some good points. I am currently writing my master thesis, so I
do not have the time to participate in any discussions right now, I
will follow the discussions though and happily implement whatever you
guys decide :-)