[OpenIndiana-discuss] from the lost to the river

James Carlson carlsonj at workingcode.com
Wed Oct 3 11:32:20 UTC 2012


David Halko wrote:
> SVR4 had a stream option, to bundle packages. Packages were often delivered
> on floppies, tapes, disks, and eventually by HTTP.

The "stream" option was really just cpio.

> SVR4 packages could be delivered in streams encapsulating 1 or more
> packages or as a bursted filesystem spooled packages. Encryption and
> compression was an option in SVR4 using class action scripts by multiple
> vendors.

It is exactly those class action scripts (and the preinstall/postinstall
scripts) that were a key part of the design process for IPS: in terms of
trying to make software from multiple vendors play nicely together and
in terms of supporting very complex systems over time (such as the Sun
patch stream sausage factory), those scripts were a serious impediment
to producing long term stability.

When everyone has an idiosyncratic device driver installer, it's very
hard to avoid undesired interactions among packages.

Some third party vendors had built entire shanty towns under those
scripting hooks -- possibly "great" for the user experience of any one
vendor's user installing a handful of packages on a single machine,
assuming that user never had to install anyone else's software ever
again.  A complete disaster if you had to do more than that.

(There were other reasons, of course.  The single central "contents"
file was an important one.  The packaging tools, which I spent some time
at Sun maintaining, spent a *lot* of CPU and I/O effort massaging that
one file.  Much of the reason that pkgadd/pkgrm were so darned slow was
groveling through that file.)

Don't get me wrong: I'm no IPS fan.  It's way over-complicated, and for
all its faults (and they were legion), SVR4 packaging was a good simple
solution.

> SVR4 packages include the ability to perform integrity checks of the
> installed package against an accepted manifest.

"pkg verify" and "pkg fix" in IPS.

-- 
James Carlson         42.703N 71.076W         <carlsonj at workingcode.com>



More information about the OpenIndiana-discuss mailing list