> Dear OpenSolaris community,
> 
I'm just a lurker, but 'hi'.

> With this post I started the work on Menhir.
> 
> Menhir is a source based distribution of 4 free
> operating systems: Slackware, FreeBSD, NetBSD and
> OpenSolaris.
> 
> Menhir will distribute the sources of:
> FreeBSD Release 6.1 (base system)

6.1 is archaic; the last release of FreeBSD is 6.2. and 6.2 is six months old. 
6.1 was released in may of 2006 which makes it painfully out of date. So 
personally, for the sake of features and security I would choose 6.2. 

Actually, for the sake of security I'd choose OpenBSD 4.1; but then you'd be 
losing features (mostly device drivers and WINE).

> Slackware 11.0 (all packages which are required for
> "starting with pkgsrc)

Slackware is forever frozen, meaning it would be hard to apply security 
patches. As much as I hate gentoo, I think they might be a better alternative 
simply because they have a mechanism for applying source-based security updates 
into the OS. 

The best choice, in my quite uninformed opinion, would be to build a skeleton 
LFS framework, and drop in packages as needed (foo-0.1a has a security hole so 
you unpack foo-0.1b into your source tree and tell people to update via 
subversion or whatever).

This might not be as unfeasible as it sounds at first; the last time I read the 
LFS site they were working on an automated LFS, so I'm sure you could pull out 
a great deal of ideas and infrastructure code from that. 

It's something to consider and I think it would be a better alternative than 
relying on slackware's packages.

> NetBSD Release 3.1 (base system)

Getting old in the tooth, but since 4.0 hasn't appeared yet, fair enough.

> OpenSolaris build 65 + two tar balls of closed
> binaries which are required.
> +

All I can say is that I hope you've had better luck building meta-pkgs/xorg on 
solaris than I have... (disclaimer, my problem may involve a lack of skill 
rather than luck).

> pkgsrc 2007Q1

2007Q2 is on the verge of being released, and 2007Q1 will no longer be 
supported. 

> 
> The base systems are _forever_ "frozen" . Only bug
> fixing and security updates will be
> included in the source files. Only the pkgsrc tree is
> a "moving" part for
> actual and fresh userland.
>  

So, are you saying that you will never update the base system? If so, I think 
you'll very quickly end up with cripplingly out-dated and irrelevant base 
systems. A much better idea would be to pick two arbitary times a year (I would 
suggest tracking when pkgsrc makes their releases) and update each base system 
then (even if it meant only distributing the latest cvs versions if they have 
not made a new release between the two pkgsrc releases).

You're doing no one any favors by distributing antiquated base systems which 
lack features and/or drivers. 

> want to distribute these 4 operating systems side by
>  side

How? On DVD? On CDs? Through an ftp site? Are you going to have each base 
system have its' own boot DVD/CD with packages, or what, exactly?

> or following reasons:
> 
> 1) Collaboration:
> I want to make Menhir the meeting point of the users
> and developers from all
> the four operating systems. They should chat, develop
> and test together their
> systems for interoperability, security, stability and
> so on.
> 
Sounds good on paper. But it also doesn't sound like anything that the pkgsrc 
developers aren't already doing.

> With this I want to get rid of the ever ongoing of
> rivalry between the users and
> developers of the (free) Unix-like operating
> systems.
> 
That goal is impractical. The rivalry is based on differences of opinion
which include both technical matters and philosophical matters (BSD license vs 
GPL
vs CDDL).

> 2) Teaching:
> I want to make Menhir a distribution for training the
> administration and development
> skills for all the four operating systems.
> You know that people are driving heterogeneous
> systems at their workplace and also
> in some cases at home. I also don't have to explain
> to you that heterogeneous systems
> are more stable and secure.
> With Menhir you will get the basis for building up
> skills for such a multi-operating system
> environment.
> 
Heterogeneous systems are a good thing; in fact that is the entire selling
point of pkgsrc.

I think you have to be careful with this, however. Someone who learns how to 
administer "Menhir" won't be at any advantage when their work places them in 
front of a redhat console. If all they know is "Menhir" they'll be lost if they 
are ever faced with trying to do techniques using sun's GUI tools.

> I witness that if today someone learns about some
> Unix-like operating system then it is
> mostly Linux. I want to change that with Menhir. I
> want to bring the broad knowledge
> about all four operating systems to the users.
> 
+1, rock on.

> 3) Experimenting:
> With Menhir you can start your own experiments
> (system building and also development) without
> fear because the stable base systems give you a point
> to where you can return if your experiments
> were not successful.
> 

But your users wouldn't really building a system, from what you're saying. If I 
install Menhir/FreeBSD I *have* the base system; what I can experiment with is 
what combination of *additional packages* I add on to it. 

"system building experiments" -to me, at least- implies things such as trying 
to get 
ulibc to work with the OpenSolaris kernel and porting the BSD userland to it. 
But if you're provided with a "frozen" base system there is no changing the 
libc, or the rc.d/smf system.

> With this I want to encourage users to build their
> systems not only with Linux but also
> with FreeBSD or NetBSD or OpenSolaris using the best
> combination for the appropriate "use-case".
> 
> The principles of Menhir are:
> I) Strength : You, as a user, get four distributions
> with one medium. 

What medium? Or do you mean from one place (ie, one internet location)?

> You, as a user, get one united
> documentation for every system. 

Are you saying that you'll be choosing ONE set of man pages to explain ALL FOUR 
systems? 

That sounds like a recipie for disaster (eg what happens when someone reads the 
linux manpage for killall and then runs killall under Menhir/Solaris)

> II) Vigour : You get a maximum flexible system. From
> the kernel up to the graphical userland you can
> choose every part of your system. As the base system
> is constant the userland will be always developing
> forward. 
How are you defining base system and how are you defining userland? 

What you're describing: including netbsd/base, freebsd/base and opensolaris 
implies 
that you would *already be* distributing individual userlands. Each one of 
those comes with fully stocked /bin:/sbin:/usr/bin:/usr/sbin, etc. 

You appear to be confusing add-on packages (pkgsrc) with userland.

> III) Security : As the base system will be constant,
> after training yourself or your employees your skills
> never get old. As the base system will be constant,
> the testing and bug fixing will make the most
> possible stable system. With the overall quality
> management of the constant base system the most
> possible secure interaction between every operating
> system will be ensured.

Until your various base systems become unsupported and security advisories and 
patches cease to be released for them. Which is what will happen if you -as you 
say- intend on keeping your base systems frozen.

> Today, I am building up the site for Menhir. Soon I
> will post you the address.
> But in the meantime every (constructive or
> non-constructive) comment is welcome.
> 
> best regards
> Gueven Bay

Debian considers themselves to be a kernel-independent Operating System. It 
would be incredibly awesome to have operating system releases which feature the 
minimum from the host os (apart from kernel, libc and OS-specific features such 
as solaris zones) and have the rest made up from pkgsrc, (eg 
pkgsrc/opensolaris, pkgsrc/freebsd, pkgsrc/linux. and so on)

The advantages I see to this are that you download your sources *once* and then 
you are able to create packages for each OS, and pkgsrc does a great job of 
keeping up with current application and desktop versions (a little better than 
OpenSolaris and MUCH better than Slackware).

I am by no means a CS person (in fact, I'm posting this from Windows XP), but 
earlier this year I spent a few months playing with pkgsrc and minimal installs 
of opensolaris, openbsd and slackware (and annoying pkgsrc-users in the 
process) with the intent of doing something similar to what you're proposing. 
So I *like* your general idea; but I think you need to reassess some goals, 
clarify some points and completely scrap the idea of having a permanently 
"Frozen" base for each OS.

Good luck, I'll be watching with much interest!
 
 
This message posted from opensolaris.org
_______________________________________________
opensolaris-discuss mailing list
[email protected]

Reply via email to