Hacker News Reader

Comment by cyberpunk | Hacker News Reader

I do know some folks that are still using it (or Joyent's SDC Hypervisor offering, which I guess also has numbered days since the Samsung aqui).

Sadly, I don't think it's got much time left. I do miss it, it was rock solid and the tooling was much better than linux at the time.

Containers were implemented properly from the beginning, SMF is (still) much superior to SystemD, things like resource limits and so on could be applied in a much nicer manner.

And it was fast too, when you had the sun compilers. Oh well.

RIP.

View on HN

Joyent SmartOS is now being maintained by MNX and the GitHub repositories have moved to https://github.com/TritonDataCenter

Oh nice. I hadn't heard of mnx.io.

They don't seem to be a huge player like Sun/Oracle, Joyent/Samsung though.

I really wish them well, I loved that stack -- but I wouldn't bet on that stack anymore. I moved storage to FreeBSD, (internet) networking to OpenBSD and VM's live on proxmox when I am running actual hardware.

Otherwise, $cloud.

Oxide has bet the farm on it.

Yeah, but oxide is Bryan Cantrills company, so I mean what do you think they were going to use, BeOS? :}

I was surprised to read recently that Haiku is now being used in some commercial products.

What products? That is surprising.

http://www.tunetrackersystems.com/products.html

TuneTracker is old school BeOS from the old days, this isn't a new development by any stretch of the imagination. They just switched to using Haiku.

It’s still Haiku being used commercially.

I don't think so. Bhyve with a Rust frontend could run on other systems.

What is the relationship between MNX and Oxide? I cant google anything on it.

I saw the MNX announcement the other day and am still confused by it. What did Samsung acquire Joyent for if not the software, or did things simply not pan out and Samsung is now offloading bit?

It felt like an acquihiring to build their internal cloud, with an hedge that if it's wildly successful, may even be offered as a public product.

However, Linux (and containers running on Linux) ate everyone's lunch.

>or did things simply not pan out and Samsung is now offloading bit?

Samsung wanted to enter the Public Cloud Space, ( The kind of business they would be interested in ), but their NAND and DRAM division would suffer order reduction from hyper scalers. In the end they sort of back off this segment of business.

There's also https://danube.cloud/ which I think is a fork of SmartOS

Zones and ZFS were great together.

It was a rock solid OS, sitting on rock solid hardware. I enjoyed working with Sun kit back in the day, and it's such a shame that Oracle killed yet another great product.

Oracle didn't kill it.

It was already dead when they bought it. What Oracle wanted when they bought Sun was Java.

Solaris and the hardware just came along with the ride. There are a few people that they needed to support for contractual reasons and a legacy install here and there that they could continue to milk, but for everybody else Solaris was dead before 2009.

This was just as obvious then as it is now. It is just that Solaris fans refused to see it.

It’s impossible to reply to this because by virtue of thinking that the systems were good you can label anyone a fan.

The fact remains: it was widely deployed, and we are still catching up to certain components of the operating system.

Dtrace, ZFS and SMF are examples.

Calling it dead only tells me that you weren’t in an industry that used it. It would be like claiming windows is dead because most things these days are cloud hosted and windows is very expensive there. It’s true in some sense but extremely blind to the rest of the industry.

Exactly, had Oracle contuned the deveopment in the open in cooperation with the community it would have been an amazing system to this day.

Dtrace, zfs and smf were nice but came too late but they are things that appeal the engineer, not the manager that sign the contracts. And Sun was already on life support when Oracle bought them.

Sure there was still a lot of companies still using mostly solaris back then - I worked for 2 of them as a Solaris "expert" somewhere during last decade - but all of them already had projects running about replacing Solaris for more cost effective redhat, even those that bought exadata racks.

"expert" meaning I was still one of those survivors still young enough to be still willing to work in an operation team and having more than 5 years of experience administrating Solaris and AIX shit.

I must say I loved the integration of zfs with gnome2's nautilus back then on opensolaris. I used it for a while as my daily driver at home.

I was actually rolling out new installs of Solaris as late as 2012.

Oracle bought Sun for MySQL. All the rest of the software assets were just icing on the cake.

And Java, surely?

Yeah it's a "why not both" situation. They wanted Java for that sweet Android lawsuit, and they wanted MySQL for the support contracts and to ensure it wouldn't encroach on many of the same use cases Oracle database is still needed for.

I think the general consensus is that Oracle bought Sun for Java.

Databases at Oracle have certainly done better than Solaris though. OracleDB is still a cash cow and MySQL I believe MySQL is still the most popular Open Source relational database ( well, other than SQLite probably ). Monty forked MySQL to MariaDB literally the day after Oracle bought Sun but MySQL remains far more popular.

> This was just as obvious then as it is now. It is just that Solaris fans refused to see it.

Solaris fans (yours truly included) saw it just fine. But Oracle has been a poor steward, and if it had any chance of being rescued, well, they gave it none. And their fame for suing and squeezing their own customers didn't help... every company ran away from it once it was an Oracle property, and suddenly there was no way to run most commercial software on it (My employer was impacted for example).

> but for everybody else Solaris was dead before 2009.

I worked in the Solaris engineering org. We had new customers at least as late as 2017.

This a thousand times.

Sun was killing Solaris way before Oracle acquired it. Because Sun was at the beginning of that era, people look at it with rosy colored glasses.

It had terrible leadership and didn't understand how to fully embrace open source (speaking strictly about OpenSolaris, they never really wanted share control of anything, form a community, etc).

I ran OpenSolaris on my laptop back in the day; I thought it worked very well. Note I'm not some old-style Unix greybeard who used Solaris back in the 80s: I was just playing around with various systems and ended up sticking with OpenSolaris.

To be fair, from my casual observations it didn't seem there was that much inertia around the system, and Oracle killed OpenSolaris before I had a chance to put it on any servers, but I definitely think there was some potential there.

> Oracle didn't kill it.

What is Solaris without SPARC? A somewhat slower and not-very-compatible OS for apps that are perfectly content with an x86?

Oracle killed it by stopping development of new SPARC systems. Linux is a much better fit for x86 boxes than Solaris ever was. Solaris for x86 was not a revenue driver or should never be considered that, but a gateway drug to SPARC workstations and servers. Java was the same - a programming environment where you could develop on Windows and deploy on Solaris. Unfortunately, for them, you could deploy to Linux as easily.

> This was just as obvious then as it is now. It is just that Solaris fans refused to see it.

Sun was having trouble funding their chip development operation because they tried to be an everything company.

If anything Oracle improved on it, by introducing SPARC ADI alongside hardware memory tagging before any other commercial UNIX.

Solaris has been a zombie for years. There's a ton of enterprises that still need it. If you suggested using it for a new project anywhere, you'd be laughed out of the room (or the zoom call, at least.)

Use omniOS, literally next best thing. Zones, smf and all the other great things. Amazing stuff.

omniOS is absolutely great for servers!!

Never heard of it so I had to look this up, OmniOS is a distribution of illumos.

And for those who dont follow closely, illumos is forked from OpenSolaris. And OpenSolaris is a discontinued open-source computer operating system based on Solaris and created by Sun Microsystems.

Exactly! OmniOS, Tribblix and Smartos are probably three most prominent ones. OpenIndiana is another one, but it is aimed at desktop computers.

> Containers were implemented properly from the beginning, SMF is (still) much superior to SystemD, things like resource limits and so on could be applied in a much nicer manner.

This! Shame they didn’t make it free sooner—it might have lived longer.

I’m curious as to why you think SMF is superior to systemd. Can you elaborate?

I'd never heard of SMF [0] before, and I'm not very familiarized with this problem domain. Maybe this is a dumb question, but if it was so great why hasn't anyone tried to replicate it for Linux?

It always baffles me that superior tools can be developed while still failing to gain traction in alternative ecosystems.

[0] https://www.oracle.com/technical-resources/articles/solaris/...

As I recall, it was a mess of XML files you edit and slow Java based tooling.

They did, it one of the larger parts of systemd that motivated its adoption and ubiquity.

A bunch of commercial UNIXes sprung init replacements/extensions with process lifetime management systems in the 2000s, systemd is visibly influenced by many of their design choices.

OS X natively uses launchd (uses plists for config, has a weak dependency system) , Solaris uses SMF (uses XML for config). There were also the daemontools/runit family, s6, and a couple others that sprung up from open source quarters.

I looked at this once, and I seem to recall SMF relies on a Solaris-specific IPC mechanism with no simple analog on Linux

https://en.wikipedia.org/wiki/Doors_(computing)

SMF, which is XML based is superior? Really? https://utcc.utoronto.ca/~cks/space/blog/solaris/SMFServiceM...

https://docs.oracle.com/cd/E23824_01/html/E21798/firstboot-2...

Yep. What's wrong with XML?

Reply on news.ycombinator.com