>
>
On Mon, Dec 29, 2014 at 2:32 PM, Vince Mulhollon <vince@???>
wrote:

>
> T.J. seems to be proposing more of a freebsd model, where xmonad is part
> of Debian but not part of Freebsd 10.1, although the UI is virtually
> identical "apt-get install something" vs "pkg install something". That
> would be a big change from the Debian way of doing things, but, it clearly
> seems to work very well for freebsd. I'm liking freebsd quite a bit over
> recent months.
>

To some degree.

I'm still willing to help out, but after giving this considerable thought,
I find myself asking what does a clone of Debian - a "one size fits all"
really accomplish? If we intend it to be a distribution for reliable
servers, then we are competing directly with RH 6.x, which RH still
maintains because not every sysadmin buys into systemd. Looking at the
proposed audience, it seems likely to be undervalued. Everyone I have ever
worked with has went with either CentOS/Scientific/RHEL over Debian, even
when I installed a Debian server at first.


What I am proposing is changing the support model. We have limited
developer talent at the present time, and that is going to be an issue, and
probably remain so until after the first release. This means we
have probably have to focus on removing systemd from the relevant packages
and copying the rest wholesale. This is not a bad idea, but over the long
term I do not think that it will be possible to maintain compatibility with
Debian and just offer a limited set of packages to offset systemd.

I think that it might be wise to just discuss - not argue - at this point
if it might be better to create a minimum core as the focus of Devuan.
This core could be updated regularly by the main Devuan team, and serve as
a basis for other groups within Devuan to extending it to desktops, servers
or whatever. It would be more manageable for our presently limited
developer resources.

I suggested that "spins" could be created for internal groups for specific
purposes: general servers, sendmail, XFCE and so on. That way our
developers could work on what they actually use and hopefully produce a
more cohesive final result than trying to create a "one size fits all
distribution." We have lots of general distributions, but where Linux has
had the most success is customized versions for a specific task. "Turnkey"
images could be created and distributed rather than simply saying "Here,
set it up yourself"