Hacker News Reader

Comment by pxc | Hacker News Reader

When Oracle bought Sun, there were a number of open-source projects they started fencing off and developing in a less community-oriented way. Usually, I think they didn't even actually change the licenses.

In every case I can remember, the community-centric forks outpaced or altogether outlived them. Hudson is dead. OpenOffice is basically irrelevant. Oracle ZFS sees little interest and is not what anyone thinks of when they hear 'ZFS'. But Jenkins, LibreOffice, and OpenZFS are all going strong many years later.

This could end up being a really good thing for Terraform users, would-be contributors, and Terraform itself as a technology. I wonder if any Terraform maintainers from HashiCorp itself will jump ship to work on it full time as OpenTF. A similar phenomenon ended up being a decisive factor with the post-Oracle forks of Sun software IIRC.

View on HN

They definitely changed one of the licenses, and in fact in the most brazen way possible -- they relicensed OpenSolaris to proprietary![0][1] Your point stands, though: illumos very much outpaced and outlived it.[2]

[0] https://www.youtube.com/watch?v=-zRN7XLCRhc

[1] https://www.youtube.com/watch?v=Zpnncakrelk

[2] http://dtrace.org/blogs/bmc/2017/09/04/the-sudden-death-and-...

I'm sure that was an especially painful change, being such a gratuitously stingy and shortsighted treatment of a dear and painstakingly refined core technology project.

My point with that observation about licensing was just that in many cases, it didn't even take something as drastic as a license change to get a project superseded by a community fork. Just changing the release process or removing some community maintainers' commit bits or whatever could be enough.

Maybe some of that was because seeing what was done to OpenSolaris made everyone else ready to jump right away as soon as they saw any shift at all in other projects, though.

Totally agreed -- and thank you for bringing up Hudson, which I have been concerned is being broadly forgotten!

I prefer to focus my imagination on picturing how angry Larry must have felt after becoming the owner of MySQL and learning that, even after spending a couple bullions on it, he still couldn't kill the thing.

And I always forget Hudson's name, but not the product.

> I prefer to focus my imagination on picturing how angry Larry must have felt

No no no, didn't you watch the video? Don't anthropomorphize Larry Ellison! Lawnmower can't feel angry!

This mention of Hudson was a blast from the past! I honestly couldn't remember whether Hudson sprang from Jenkins and failed to gain mindshare or whether Jenkins sprang from Hudson and took all the mindshare. We used both of these on a project I working on from '08 to '13, and I guess we must have started out on Hudson and then migrated to Jenkins. So you're right to be concerned about peoples' memories fading!

(I honestly thought both projects worked very poorly, but also better than anything else we could find at the time.)

> (I honestly thought both projects worked very poorly, but also better than anything else we could find at the time.)

Each of us must first earn the right to forget Hudson by finally decommissioning our respective companies' remaining, bespoke, janky Jenkins server. ;)

hahaha I took the coward's way out and left that company a decade ago. When I met with an old colleague last year, they were still on that bespoke, janky Jenkins server.

First time caller, long time listener:

You have a blog too! Hurrah! I celebrate your entire catalog. If you start another podcast, please be sure to mention it on oxide and friends. It was crickets over there in the "On the Metal" podcast, but when you guys made that last episode I had 2 years of oxide & friends, so perhaps that was ok.

Definitely no third podcast in our future! Oxide and Friends itself wasn't even all that deliberate -- it was very much an artifact of the pandemic[0]. Glad you found Oxide and Friends, and sorry it took us so long to record an On the Metal episode to point to it...

[0] https://www.youtube.com/watch?v=W8qiDhlFVCE

Sun is a great example of a company that had excellent engineers but a terrible business model. Instead of fixing the business model, they tried to extract as much value out what they had and it ultimately lead to the death of the company.

Terraform is the gateway to Hashicorp and it feels like they're going to lose it.

I was just a kid when Sun died, reading about all the drama of the buyout and the series of forks on Slashdot whenever I had already finished my work in my high school computer science class every morning.

I didn't really know what or how to think about the mismanagement aspect of it, but I was already a F/OSS enthusiast by that time. I ran Linux at home and used free software everywhere I could. I even eschewed Times New Roman in favor of Linux Libertine (then the font of the W in the Wikipedia logo!) for my school essays because I didn't want to support the hegemony of a proprietary typeface.

What I did understand about Sun was that they had been the caretakers (and sometimes creators) of some of the most important and beloved technology I'd ever used. I didn't know anything about Solaris, Hudson, ZFS or SPARC, but I knew Sun for VirtualBox, OpenOffice, the MySQL database system that seemed to be everywhere in the open-source world, and the Java programming language I was writing in for school.

I remember a sense that something precious had died with that buyout, even though I didn't understand what led up to it. I worried a lot about OpenOffice, which I relied on and advocated to friends, and I practically cheered when I read about the formation of The Document Foundation and LibreOffice.

Yes. Sun became all about extracting rents from vendor lock-in (SPARC, J2ME, SUN DS, etc.). But Sun couldn't maintain that vendor lock-in, so Sun went under. Management just did not or would not see this, much less act on it. OpenSolaris was a good initiative for rebuilding mind-share, but it was a bit too little too late.

But the biggest mistakes Sun made were made in 2002-2004, and they never recovered:

  - the Solaris on x86 "cancellation" (it
    was more putting it on hold, maybe, but
    the market took it as a cancellation)
  - not making a deal with Google for Google
    to use Solaris in their data centers
  - killing off Sun PS (professions services)

The Solaris on x86 "cancellation" killed a lot of mind-share. No one wanted to get locked-in to SPARC.

Making a deal with Google would have reinvigorated Solaris' mind-share at a critical time.

IBM did the opposite of killing off its professional services division. IBM killed off their x86 systems division and beefed up its PS, and IBM went on to make a killing on PS.

The vendor lock-in issues totally compounded these terrible decisions.

After 2002 Sun was always on the back foot.

> not making a deal with Google for Google to use Solaris in their data centers

Wow i had no idea that had almost happened!

Anyone care to elaborate further on that?

I wonder what could have been, if later solaris was open sourced and google could push their improvements… maybe we’d all be running containers based on solaris zones, and would have solaris on our smartphones? Who knows!

The story as I understand it is that Sun wanted to license Solaris on a number of hosts basis while Google considered the number of servers they had to be a critical secret.

> maybe we’d all be running containers based on solaris zones

There were a lot of nice things about Solaris. The Solaris engineering org was amazing, full of visionary giants. ZFS, Zones, DTrace, mdb, SMF (systemd, but done better and ten years earlier), the fault tolerance stuff -- all created by engineers as skunkworks.

Fun fact - some Google servers ran Solaris. I worked on the tape backup team and the machines connected to LTE drives were indeed running Solaris.

TF's excellent plugin ecosystem that is not owned by Hashicorp will teach them a bitter lesson.

On the other hand, for many years they were maintaining most of everyone's plugins on their own repos. It's not like Hashicorp hasn't been an honest servant of the community.

I understand people's dislike of the license change, but Hashicorp isn't Oracle.

Not really. I was a frequent contributor to plugins and it was actually incredibly frustrating getting changes merged for many of them. Many you couldn’t test your changes without licenses that cost hundreds of dollars. They controlled the repos, but they didn’t maintain them.

It's becoming it though.

Not really. Oracle provides software that does not require any activation cose but will phone home and snitch on you so that oracle can send you threats if you don’t pay license fees.

I’m thinking of some additional virtualbox guest additions.

> but will phone home and snitch on you

Terraform would never do that, would it? Oh, wait... [0]

[0] https://github.com/hashicorp/terraform/blob/v1.5.6/checkpoin...

It's somewhat ironic though, that the license adopted by Hashicorp, the BSL, originates from MariaDB, which is the no-oracle successor to MySQL.

And arguably MySQL does better than Mariadb nowadays.

I think the only thing that wasn't forked off from the Sun deal was VirtualBox, mainly because Oracle never cared much about it and it's still open source, and partly because the container revolution was already brewing.

Reply on news.ycombinator.com