Remove BIND, Sendmail, Perl and etc from base?

Wesley Moxam wmoxam at klickit.com
Thu Jul 24 13:12:47 PDT 2003


----- Original Message ----- 
From: "Chris Pressey" <cpressey at xxxxxxxxxxxxx>
Sent: Thursday, July 24, 2003 3:32 PM
Subject: Re: Remove BIND, Sendmail, Perl and etc from base?


> On Thu, 24 Jul 2003 20:45:33 +0200
> "Simon 'corecode' Schubert" <corecode at xxxxxxxxxxxx> wrote:
>
> > should the packaging system be completely contained in the base
> > system?
>
> Well - if it's not in the base system, then it's ... um ... a package :)
> How is that circularity resolved?  Some sort of bootstrapping?
> Or is it just a simple tarball?  That seems inelegant, somehow.
> I think, on balance, I'd prefer for it to be in the base system.

Existing package systems are available as a tarball ... it really isn't a
problem. The system could alternately be installed during the OS install
process if desired.

> > pure shell scripts do the job very well too.
>
> I don't have much against 'make' (I have more objections with the *way*
> it's used, than the tool itself.)  But I could see shell scripts doing
> the job just as well, assuming they're generated by some tool which
> knows the dependencies and does the topological sort.  (That tool could
> be a perl/whatever script, run by the package author, and the end user
> installing the package wouldn't need perl/whatever installed.)
>
> But also, considering that many, many ports come with their own (usually
> GNU) makefile anyway, the 'make' approach might not be worth throwing
> out.

Is an entirely new packaging system really needed? This seems to be a
entirely seperate project in itself. There are already several mature (and
not so mature) package systems that have been tried and tested, and they
already have many packages available for them. For example, NetBSD's pkgsrc
contains nearly 4000 ports and is supported on 7 operating systems.

It will take a lot of time and effort for a new package system to have the
features and number of ports as these current packages do, not to mention
the effort it takes to keep ports up to date. Wouldn't it be more reasonable
to start with and existing system (like pkgsrc), and then add features as
needed? Just a thought ...

-- Wes








More information about the Kernel mailing list