On Sat, Dec 6, 2014 at 2:24 PM, Isaac Dunham <ibid.ag@???> wrote:

> It seems to me that there are several propositions that have been advanced
> for arrangement of repositories other than "main".
>

With all due respect ethe schemes are putting the cart before the horse.

Talk about WHY you want to segment repos, before debating schemes.

1) In the olden days there was non-US because USA wasn't as free as the
rest of the world WRT crypto software.

2) The idea of main exists because I have personally printed out the DFSG
and a copy of my /etc/apt/sources.list showing I use nothing but main and
handed it into the licensing hired guns and (successfully) told them to
pound sand. You should have seen the all nighters the windows admins went
thru... and their six figure licensing bill, LOL. Also every mirror
operator in the known world, AFAIK, can mirror main without any concern
about licensing, although if I were operating a public mirror I'd be
nervous about mirroring non-free... what if there's one non-free package
that says I can't redistribute because I'm ($whatever) so it causes all
kinds of legal probs... safer to just use and mirror main only.

3) The idea of contrib was its free software thats useless without non-free
related items. Its never been perfectly enforced. Technically the kernel
belongs in contrib due to non-free firmwares and the PS/2 driver in the
kernel is useless without the non-free firmware embedded inside my
trackball. In an era of 900 packages and 4 meg 40 MHz 386 (my first linux
box in oct '93-ish running SLS), it was probably necessary to split to make
"main" a little faster. In an era of 10K packages and octo-core 4 GHZ
machines or even 700 MHz raspberry pi, its not really necessary anymore.

None of those use cases seem to have changed other than #1 being legally
sorta-obsolete but not entirely (someone needs to look into it, #3 making a
distinction between contrib and main is no longer computationally
significant.

Once the use cases are defined, the resulting scheme will likely be very
self evident. Unless someone has a use case that hasn't been mentioned, in
2014 we probably need main and non-free.

Debian has a horrific problem across numerous areas, technical, procedural,
and social, of being stuck in the 90s, and a clean fork-ish start is a good
time to clean shop. So bye bye contrib, you were computationally necessary
in '96 but not today.