I will always look back on Solaris wistfully. It was (and will always be) my favorite Unix. When containers showed up... oh man... my lab guys lost their mind. We had a never ending parade of upgrades...and server rotation. Enter zones... and things just become OOM easier. Want a zone of your own? No prob. Wonderful isolation and resource allocation, too. The network stack was a bit iffy in the early days, but Sun sorted it out. All our zones (incl. root!) were on NAS... we could move zones around the hardware at will. It was incredibly useful and a whole lot of fun.
Solaris 10 really was the pinnacle of UNIX. At a previous company, we were doing with Solaris zones and ZFS snapshots essentially a docker-like development ecosystem. We were really excited about the crossbow networking project, but then the Oracle acquisition happened almost all new and innovative work stopped.
In a more perfect world, Joyent would have become bigger than it was. Alas...
I'll assume that was a typo and you meant Solaris 11. Solaris 11 has some amazing advancements over 10 in terms of security, packaging, and many other metrics.
Also, for the record, it is objectively untrue that all new and innovative work stopped. My primary disappointment was the silent discontinuation of the OpenSolaris project.
No, I meant 10. The same way you might say that the McLaren F1 is the pinnacle of car engineering, despite higher performing cars being subsequently released.
The delta between what you could get between Solaris 10 and
most anything else at the time (free or otherwise) was very impressive. Zones/containers had little equal (maybe FreeBSD jails, but so much nicer to manage), ZFS made me almost cry that I’d never have to deal with veritas again, and DTrace allowed safe, live debugging whenever I felt like it. I’d add SMF there somewhere if it wasn’t configured via XML.
Sure, IPS is technically superior to apt or yum (well, anything is better than yum), but it’s not ground breaking better than what you have elsewhere. 11’s security is indeed improved as they took what was already done of crossbow, and implemented it, then cancelled the rest.
I was using hyperbole, but the general feeling to those of us that were customers was of a large cutback in work. And that was before oracle jacked up the prices.
I miss Solaris, but there is no reason I would recommend it to anybody today, even if you had to manage your own hardware and not be on a major cloud vendor.
Then we're going to have to agree to disagree; because the next Solaris release went all the way to 11 ;)
As for 11’s security is indeed improved as they took what was already done of crossbow, and implemented it, then cancelled the rest. -- that's shortchanging an incredible amount of work that was done that was not related to crossbow at all. Crossbow, while great, was hardly the largest or most important project that was being worked on although it was certainly one of them.
For me, the biggest shame is that Linux never got to adopt all these brilliant technologies like ZFS or Dtrace, directing efforts to alternatives that are nowhere near the quality or reliability of the former.
Yes I know there's a long time debate about whether the CDDL license is GPL compatible or not. Too bad this came from the Sun era, years before the Oracle acquisition.
The problem was they grew hugely during the dot com bubble - to a size that was never sustainable, and then post bubble, couldn't work out how to make enough money from their offerings in the face of Linux on commodity PC hardware.
Not just OS/hardware either - the Java ecosystem and Libre/Open office still live on.
Poking around in the weirdness that is Solaris is so much fun. The comments in the source code are legendary as are the man pages, seemingly much more helpful than other platforms, with a focus on providing good examples. Zones, DTrace, ZFS, and SMF are incredible tools that were over a decade ahead of their time. Zones in particular seem to still offer isolation advantages over the various implementations of containers on Linux. Designing a cohesive system for containing multitenant workloads had its advantages. I'm glad enthusiasts have been able to keep Illumos going, it would be a shame for nobody to be running all that code.
Asserting that Solaris has weirdness is an insult, because Solaris is the AT&T System V Release 4.0 reference implementation. Solaris is also the reference implementation for libc functions like malloc, threading, realtime kernel implementation, NFS, NIS, containers, shared memory, SCSI, fiberchannel, high performance TCP/IP networking, network virtualization, parallel service startup/shutdown, POSIX compliant shells, POSIX AWK, XPG4 and XPG6 userland, I could go on and on and on; if you want to write an implementation of some IEEE, POSIX or RFC specification, or some kernel or userland subsystem, Solaris / illumos is the place to refer to on how to do that, and how to do it correctly.
Some comments are treating this as an opportunity for a Solaris wake and while I certainly don't object I wonder if needs pointing out that there is a lively ecosystem of Illumos derivatives, Illumos itself being a derivative of Solaris (via what was once called OpenSolaris).
Some examples include SmartOS (Samsung maintained datacenter centric flavot, lots of tools for rapid/easy deployment of compute nodes and machine images); OmniOS (for laptops); and one I had never heard of until this post, Tribblix.
A common pattern is to run other OSes in some zones via bhyve (linux and bsd) or kvm or lx branded zones (linux). But the native platform can handle many scenarios these days.
Just a note to say Tribblix is a fantastic Illumos distribution that works well on AWS. Peter Tribble deserves much applause for maintaining this. If you are curious re: Illumos and need a place to start - Tribblix is a good one. Have fun.
Solaris 10 has been EOL since what this time last year? No more updates are going out to it. So if you are still running a Solaris 10 zone then you aren't able to patch it.
Tribblix is unlike Slackware - it uses AT&T SVR4 packaging with zap which is akin to yum on top, whereas Slackware used tape archives. Yes, tape archives for backing up to tape. That says it all.
No pardon - Slackware "packages" were .tar tape archives. The entire "OS" was hacked together out of parts, whereas Peter system engineers Tribblix. No pardon.
Just because "tar" happens to stand for "Tape ARchive" doesn't mean tarballs have anything to do with actual tape archives. It only means that the files are stored sequentially, which is a perfectly reasonable way to structure an archive.
If you want to criticize how Slackware does things, there are myriad better points you could've brought up, like:
- The installer being only minimally changed from that of Softlanding Linux System
- The installer still operating under the notion of "disk sets", from back when it was remotely feasible to install Slackware from floppies (which hasn't been the case in multiple decades now)
- The lack of dependency checking (though having been burned by dependency hell multiple times, I'd call this a "feature")
- The lack of PAM, if that matters to you (though the -current branch now includes it, so Slackware 15.0 probably will, too)
And yet, none of these things have prevented me from being happy and productive with Slackware, both at home and at work, on desktops and servers. It's what I'm running on the very laptop on which I'm typing this comment, on my main workstation / gaming rig at home, and previously on my servers (before I acquired a taste for OpenBSD and SmartOS).
Yes and no - as they lost the University crowd a while ago to Linux, you are pretty much on your own if you want to compile some third party code on Solaris, so while the OS is perfectly viable on a subset of hardware - it's not something you'd either want to install on a laptop or try and build and run random code from github.
