Re: testing, testing
Good post, Brent. Not at all too long...
According to Brent Fulgham:
> Although I don't represent Debian by any stretch, I am going to argue the
> Debian perspective for a moment.
Same for me on the BSD side. I'm not part of the Core
group; just a porter and user who requires a rock-solid
stable environment.
>
> We (Debian) need to be asking what project format makes the most sense in
> terms of our resources, and the needs of our "users". Frankly, the amount
> of effort involved in doing some of the things Gary is talking about are
> rather large. However, that doesn't mean that they should not be done.
>
> Personally, I am of two minds on the BSD port, and I would like to hear
> comments from both sides (or more sides if they exist) as to the pros and
> cons:
I think we are approaching this from like viewpoints. I think
the user community would greatly benefit from merging all the
``linux-goodies'' with the BSD kernel+networking.
As more people joined this debianBSD effort, the work would
proceed that much faster. Realistically, I don't think we
would have much of anything to alpha-demo before several months.
In a year or so, yes, probably.
Also:: this doesn't have to|need to happen in one chunk; it
can be staged; it can evolve.
>
> Options:
And some feedback:
>
> 1. Provide a Debian "BSD Kernel" package.
> The obvious advantage here is that we drastically limit our scope to
> producing a package that can co-exist with an existing Debian GNU/Linux
> installation and provide a BSD kernel. I'm not really sure what the
> advantages over a Linux Kernel distribution would be. We would be relying
> on BSD's Linux-compatibility layers, which may or may not be perfect (I have
> no personal knowledge). We would have limited project scope, and a limite
> amount of resources needed.
>
This might be the best way to get off the ground. I think
that the FBSD Linux-compat code is very well done. So far,
everything that has been in the FreeBSD port tree from
Linux has worked. Things like WordPerfect8 ...
But as far as I know, no drivers have been ported. Because
the Linux kernel seems to be too different, nothing that
touches the Linux kernel has been attempted.
I have utterly no idea what's involved here.
...
> 2. Provide a Debian BSD "port".
> Here I mean "port" in the Debian sense. This would be similar to our effort
> to port the Debian user environment to the GNU/Hurd. I have done some work
> on the Hurd project, and we definitely get to spend time debugging more
> esoteric things than a standard Linux package project does. However, one
> thing that hurts the Hurd port is the lack of qualified persons helping.
(A few years ago I was drooling at the idea of running Hurd;
glad to hear it is still alive, if bearly... . )
> The BSD port will have the advantage of a potentially large group of
> knowledgeable individuals (the BSD folks themselves) who could provide
> assistance. But why produce this port? The main advantage I see for a BSD
> user is that they would have access to the Debian package system, which
> might make it easier for someone to use BSD, although I'm not convinced that
> there is much difference in the "degree of difficulty" in setting
> up/maintaining a BSD system. Gary points out that there are many Linux
> device drivers and libraries that are not necessarily available on BSD.
> This would obviously be a lot of work, but is potentially a very satisfying
> and lasting contribution to the Free Software community.
The beneficiaries would be the users. There is a GNU|Sys5-style
rc package for FBSD that lies largely unused because the FBSD
Core group didn't like it. I understand that there are sysadmin
packages for Linux that use a GUI (?) And I'm sure there are
many other GNU//Linux tools that would benefit the BSD side.
I'm sure this cross-pollination works the other way, of course.
Especially in the networking arena....
This would be a huge win for the FreeNix//open-source//Free
Software movement. Endless stability plus a great selection
of programs. Easier to configure and administer... .
>
> Potential Pitfalls:
>
> License issues. Debian is heavily GNU-influenced. BSD folks who are
> joining this list should probably take a look at the Debian Free Software
> Guidelines, and our Constitution. We may have some differences of opinion
> over how things are licensed. And we certainly (Debian) do not have a right
> to take BSD code and relicense it wily-nily. We are obliged to honor the
> BSD terms of the original code. The BSD license is completely compatibly
> with the Debian goals, but I am not sure if some the GPL'd kernel code is
> "safe" to port to BSD. Anyone who can speak knowledgably on this issue?
This issue can nearly raise the dead to argue, fight, and
throw spitballs. My understanding is limited to thinking that
the Berkeley license is virtually 100% open. You've got to
keep the BSD license clause on every file that carries it,
but that is as far as my knowledge goes.
I'm well into my third year of hacking a light-sound X11
``mind-machine'' application. Alpha (or pre-alpha) release
is sometime late this fall. I'm going to use the GPL when
I publish this.
In other words, I have nothing against the Copyleft.
>
> Perhaps a BSD expert (Gary?) could give us Linux folks some pros/cons about
> BSD? I've installed it on one machine a long time ago (before starting to
> work with Linux). I haven't really worked with it much since then, but look
> forward to learning more about it.
>
Disclaimer: I'm not kernel or network savvy. I did go to
Cal, but this was in the early days of BSD-4.X and I never
got into the 4.X code itself.
I'm been using FreeBSD since the early days (v 2.0.5) and
can give you my opinion as a user. BSD really is rock-solid.
It is widely known that FreeBSD stands up to extremely heavy
loads, but that won't have much impact upon most users.
I use my home systems mercilessly, am up always (I have
twin machines). In four years I've had exactly 3 crashed
that were probably kernel-related.
If you don't need this much stability, then consider that
the networking code is flawless.
Nutshell: BSD is better than Linux in its networking and
kernel stability. I'm sure there are others who could go
on for paragraphs and accurately.
Anybody else?
gary
>
--
Gary D. Kline kline@tao.thought.org Public service Unix
Reply to: