Wandering Thoughts archives

OpenSolaris versus Solaris

Discussions about Solaris often wind up bringing up OpenSolaris (and sometimes it crops up in other contexts). However, I do not find OpenSolaris particularly interesting.

My view is that right now, OpenSolaris is essentially a technology demonstrator. You can't use OpenSolaris by itself, because it lacks sufficient support to use it directly and sufficient genuine openness for sensible people to base stable things on it. You can't use it as the future of Solaris, because I'm not aware of any commitment from Sun that OpenSolaris (at some point) will be Solaris 11; instead I expect them to pick and choose elements of OpenSolaris, whenever they decide that it's time to create Solaris 11. So if you like something in OpenSolaris, it's possible that it will show up in Solaris but quite possible that it won't.

(The Oracle acquisition throws a great deal of uncertainty in all of this, too.)

As a technology demonstrator, OpenSolaris is interesting to people who care enough about where Sun is going to peer at the the tea leaves in an attempt to learn something, anything, about what's coming. I do not fall in this category; we are not that invested in Sun, despite appearances.

Now, this is in some ways a very pessimistic views, as there are specific pieces of OpenSolaris that regularly migrate to Solaris proper. For example, ZFS changes move over on a reasonably regular schedule, so OpenSolaris is a decent way to get a preview of what you will be able to do in Solaris within a year or two. However, for larger scale changes (such as fixing Solaris's packaging system) I feel that this definitely applies.

The shorter version: OpenSolaris is to Solaris as Fedora is to Red Hat Enterprise Linux, except that OpenSolaris is not as useful or as interesting as Fedora.

Thus, 'fixed in OpenSolaris' is for me not very much of a positive development with things that are wrong in Solaris. Things in OpenSolaris may or may not go anywhere that's actually interesting to me, and they'll only be real solutions when they're in Solaris itself.

Why the Solaris packaging system is bad

I recently read this holyhandgrenade.org entry, which rises to the defense of the Solaris packaging system on the grounds that it's just misunderstood; on Solaris, unlike elsewhere, the packaging system is intended only for system components (and Solaris defines this narrowly), not additional software.

I disagree. The reasons I dislike the Solaris packaging system have nothing to do with how widely used it is (and how much software doesn't come with Solaris); I dislike it because it is, purely and simply, a bad packaging system.

The issues I have with it include:

• portions of it are remarkably slow.
• it's not clear to me if it has an idea of upgrading packages (as opposed to removing and then re-adding them).
• it lacks any sort of robust signature verification. sum is not sufficient even to be confidant that you're detecting accidental damage.
• it does not have robust, end to end dependency handling, which is handily demonstrated by Sun's ongoing habit of releasing patches that are broken because of missing dependencies. Enforcing listed dependencies at install time is pointless if there is nothing that makes sure that those dependencies are correct at creation time.
• it lacks a robust set of commands to do basic package querying operations: what's on my system (considering only Solaris itself, not third-party software), what files are provided by something, what thing provides this file, and so on.

For people who think that Solaris actually has the latter, please tell me how to discover where my /usr/lib/libzpool.so.1 came from (as a hint, the answer is not the SUNWzfsu package and no, we have not installed a locally compiled version).

(I personally think pkgadd's interface is irritatingly broken by default, but I have a violent reaction to programs that insist on asking me stupid questions, and another violent reaction to ones that spew pointless information.)

Now, you may argue that some of this is actually a fault in the patch management system, not the package management system. I reject that argument, because patch management is part of package management. That Solaris has a bad patch system that it has not integrated into package management in a meaningful way is a significant part of why Solaris's overall package management is bad.

(Bad patch management is not the only problem that Solaris package management has; it would still be far behind both pleasant and the state of the art even if patch management was perfectly integrated. But it would at least be usable.)