Implications of pkgng, was Re: Why Are You NOT Using FreeBSD ?

[Matthew Seaman]

On 03/06/2012 02:21, Erich wrote:
>> 2. No decent packet manager (I hope pkgng will make life easier). You
>> > can't just upgrade this and that packet and see what's new, and
>> > rollback if you don't like somthing .

> I really hope this will never come. Why? It will kill make install.
> Make install is the key to FreeBSD.

I doubt very much indeed that pkgng will "kill make install."  What it
will do is fill in a widely recognised gap in FreeBSD's offering: that
FreeBSD is essentially unmanageable at the moment only using binary
packages.

While compiling from ports really is the gold standard for tunability,
configurability and lots of other good things, it lacks quite a lot in
terms of speed and convenience for most users.  This is something that
turns off lots of new users coming from Linux or MacOSX before they have
really had a chance to get to grips with the OS and start to appreciate
its finer qualities.  Somewhere that FreeBSD does itself absolutely no
favours.

The ideal which I think is attainable with pkgng is to be able to use
the ports to compile your own customized versions of the applications
which are critical to you, but otherwise rely on pre-compiled packages
for anything else.

Also, consider people developing embedded devices (and this is going to
be a *major* area for FreeBSD in the future) -- you want to compile
software offline (probably cross-compiling on a completely different
architecture), strip out inessential documentation and so forth and
manage precisely what is installed on your device using an efficient
binary package manager.  pkgng promises to make this feasible.

PC-BSD's .pbi package format is another way of addressing this same sort
of problem, and although originally aimed at desktop users it makes a
lot of sense for server side usage too.  As .pbi and pkgng are actually
complimentary, rather than in competition, expect some interesting
developments in the relatively near future.

> I believe a better solution would be versioning of the ports tree.
> When the ports tree compiles fully, it can be saved and its version
> number incremented.

This is simply a non-starter.  Remember what the ports tree is: *third
party software* ported to run on FreeBSD.  That 3rd party software is
going to continue to be developed at its own pace without any reference
to the FreeBSD project.  You can't just choose a point in time for the
ports tree and expect it to keep working effectively.  Particularly not
if you admit the possibility of updating some packages for security
reasons or other egregious bugs.  Nor will most users be content with
running year-or-more old versions of most software packages.

In fact, one of the big plusses of the ports is how up to date it
generally is.  Most big packages are updated in the ports within days of
updates being published upstream.

> I do not believe that much more would be needed. Of course, we have
> then a huge number of versions. Would it matter? Give the ports tree
> the major version number of the latest release. So, at the moment it
> would be 10. Increment then the minor every hour if you want. Just
> make sure that the ports tree can be downloaded for some time under
> this version number.

What exactly is this supposed to solve?  Simply attaching a number to
the ports tree won't do anything.  There is already a promise that the
ports should work on all supported FreeBSD release branches.   Besides,
with the imminent switch of the ports from CVS to SVN, there will soon
be a global revision number for the whole ports tree.

	Cheers,

	Matthew

-- 
Dr Matthew J Seaman MA, D.Phil.
PGP: http://www.infracaninophile.co.uk/pgpkey


-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 267 bytes
Desc: OpenPGP digital signature
Url : http://lists.freebsd.org/pipermail/freebsd-stable/attachments/20120603/b26aa330/signature.pgp