Goodbye Joyent
zinascii.com
zinascii.com
Really happy Z got a chance to work at Joyent though, even if it was only for a few years.
For most of us ex-Basho (ie. Riak) engineers, Basho was a career highlight. We've mostly moved on to better paying, more stable jobs, but lost the magic / passion / joy that we had at Basho.
Ryan's the exception. I imagine he's loved his time at Joyent as much, if not more, than Basho.
Here's hoping lightning strikes for a third time and Ryan lands somewhere great after his winter break.
He wasn't exaggerating, that is an epic, beautiful, I would even say humbling block comment to read. It's completely outside of my regular knowledge domain but that is not at all required to appreciate it.
[0] https://github.com/joyent/illumos-joyent/blob/dff0473c792fc5...
software would be soooo much better if we lived in a world where a) engineering practice demanded devs document this well, and b) the majority of devs had the skill to write so well. market pressure however demands faster delivery times, and who cares about the tech debt.
Is Samsung doing nothing with joyent? Seems they could and should be competing heavily in the server space if not the public cloud itself more meaningfully than they seem to be today.
Presumably not for Ben Noordhuis: https://www.joyent.com/blog/the-power-of-a-pronoun
That blog post is why I resolved never to do business with Bryan Cantrill or Joyent. I don't know if it influenced others to do likewise, but I can't imagine that it actually helped the company succeed.
I can't believe he would actually post something like that on Joyent's domain.
It's a personal attack on someone doing something he disagrees with. Shaming via a public space like that is just not done. It's okay to agree to disagree on a subject like gendered pronouns, just like it's okay to disagree on abortion laws or gun control.
NOTE: This isn’t meant as an insult; it’s just that often people use terms as if they were universally known. He had hyperlinks to some of the more obscure items on the list but not the key items.
Linux has come a long way since, which is a large part of the reason why SmartOS has become less relevant. The latter being a great shame because competition breeds innovation and we are losing a lot of interesting interesting Unixes from the public consciousness.
Edit: oh come on. I post this and it almost immediately gets negative karma despite being both factual and informative. A perfect example of the rife abuse of peer moderation on this site. I honestly don’t think I’ll bother wasting my time on here any more.
As for the element of surprise, not everybody moves in the same circles. You might be deep into server-grade operating systems, or DevOps, or other branches of information technology that are far from my daily experience. I haunt the areas associated with theoretical computer science, applied mathematics, finance, and the ins-and-outs of the (European) payments system. Each of our haunts are so vast that it’s easy to be erudite in general and yet still pretty ignorant of things others take for granted.
Namely RTOS, Zephyr, NuttX, mbed, and Arduino compatible bare metal implementations.
Common to all of them? None of them is GPL licensed.
Back in 2014, Clojure-based web development had all kinds of groundbreaking ideas that massively improved quality of life: immutable data structures, hot code reloading (figwheel), decoupling of state and view (devcards), with all the benefits that come with functional programming.
The JavaScript ecosystem has since closed the gap and has a lot of great libraries or tools that do these things, and it's just _so popular_ that it's hard to make the business case for Clojure(Script).
Personally, I still think Clojure has a lot of really smart people contributing to it. It's exciting to see tools like libpython-clj enable Carin Meier's work around leveraging bleeding-edge Python Machine Learning libraries in Clojure, and something in my gut tells me that Rich and Co. are on the right track with spec while the JS community pursues TypeScript instead.
It's true that every project needs a community if it's going to survive, and because the internet makes it so easy to share ideas, being the birthplace of good ideas is not enough to ensure your survival.
I'll be interested to see what the landscape looks like in another half a decade!
Unfortunately, when it comes to hardware support, only the most popular OSes actually thrive :(
Since the kernel and coreutils are GPL there aren't any sue-happy lawyers waiting to pounce upon CEOs saying "We use Linux", the broad family of Linux kernel based OSes (Debian, RHEL, Ubuntu, etc.) have generally gained the marketing term Linux.
I feel that Solaris missed out on this in a shockingly bad manner. Perhaps ex-Sun/Joyent folks can correct me if they feel I'm wrong, but it seems that the Solaris "family" of operating systems has always been way harder to understand than Linux-based, mostly open-source systems. Do I want illumos, OpenSolaris, or OpenIndiana? Or do I actually want Oracle Solaris? Which is the one that Oracle will not try to sue me for?
So here is a broader lesson for people trying to open source their OS, if there is ever a grand project like Solaris. Make it all open source at once, and use a copyleft license like the GPL. Otherwise you risk this kind of breakdown where there are N different operating systems that are vaguely related to and worked on by the same people, but you can't name them in a coherent manner or rely upon broad similarities.
If Solaris source had always been available under the GPL, it would have removed the incentive for Oracle-Sun to sue other people for announcing support for OpenSolaris or RedSolaris, or whatever. Of course they might have done that anyway, but in 2020 they would have been more likely to donate it to the Apache Foundation or something. Or we could have had the competing Solaris Foundation, promoting its Sun-centric technologies like dtrace, zones, and zfs as first-class alternatives to the corresponding Linux technologies.
Obvious disclaimers: Not a lawyer and not necessary good at predicting alternate futures :)
The CLA is actually what initially prompted the illumos fork even before Oracle closed the gate.
Joyent initially had a CLA on Node.js for business reasons that (as far as I know) everyone in engineering disagreed with. When we were finally able to make Triton (née SmartDataCenter) open source we also eliminated the CLA for node.
We now have contributions from many people under the MPLv2 in Triton, and we are no longer the exclusive copyright holder which means that it is pretty much impossible\* for Samsung to close it again.
* We would have to either rip out all those commits or get every contributor to either relicense or assign copyright to Joyent.
I used some of their Solaris VM's and never had a problem getting eased in. Learning how to get around in one seemed to be just a bit more work than learning a different Linux distro.
The company was fun to be a customer of even if I did lose my Mixed Grill, later upgraded to the 3 Martini Lunch. I imagine even more fun to work for.
I was surprised at how appropriate and familiar this phrasing felt.
Never heard of the OS mentioned
I hope for your sake that this admission won’t cause you to be brutally downvoted too. Here’s a pre-emptive +1 ;)
What is wrong with the denizens here nowadays??
It’s doubly tricky, because most companies pay lip service to passion. But like a chef that cares a bit too much about which cookware you use, it sometimes hinders the goal of serving 100 customers at noon.
I respect that they’ve created their entire networking stack from the ground up. Few can claim the same. But objectively, the less eyeballs on a piece of code, the more risk. What will the company do if they’re hit by a bus?
I’m not sure that companies should be measured in terms of “what if a dev suddenly goes away,” since creative work isn’t interchangeable like gears in a machine (nor would we want it to be). But it seems at least partly relevant.
The tradeoff is that if you use unpopular software, like lisp, you can often gain more leverage in specific situations. HN’s software is the most flexible codebase I’ve ever seen, but few have ever studied it deeply. Viaweb used those principles to dominate the competition at a time when everybody was writing websites in C. But are those days gone?
It might be an asset for a startup to use obscure software, so it’s definitely worth honing the skillset. I’m just not sure it’s a good idea to feel so strongly about business in particular. It has a way of burning you.
In Paul Graham's days, Lisp was not as unpopular as it is today. It was more like choosing to use Ruby in 2020. On its way out but still in the public consciousness.
It was started in 1995 and acquired in 1998, so that wasn't a very long time for lisp to come into the public consciousness. It seems like at best it was considered an AI research language.
> It was started in 1995 and acquired in 1998, so that wasn't a very long time for lisp to come into the public consciousness
Lisp was the only game in town other than C for multiple decades. Absolutely everyone learned it in university. It was the main garbage collected scripting language, like Python today.
And when Viaweb came to be, there was also ML, Smalltalk, Perl and Tcl to choose from.
This if we focus only on the context of UNIX, because otherwise the language spectrum gets even wider.
I think I've also seen GUI apps exported over HTTP using clickable image maps and early tricks for streaming images.
Viaweb was one of the first "make your own shop" web applications, yes. Not first web application.
Seriously, the AutoCAD community has some of the brightest (or, at least, most clever) developers I've ever met.
There should be an active policy of reduction information siloing that is active all the time, especially for a big corporation like that.
It puts me in mind of a line from the movie Gettysburg (https://www.imdb.com/title/tt0107007/?ref_=tt_ch). Robert E. Lee, explaining his decision to make a risky attack to one of his lieutenants, says “to be a good soldier, you must love the army. To be a good commander, you must be willing to order the death of the thing you love.”
A business built around a particularly clever bit of code faces the same dilemma. To be a good programmer, you must love that code. But to be a good businessperson, you must be willing to order the death of the thing you love.
While you have a point in general, I don't think that was a primary reason for Joyent losing the "cloud wars". At the most you can say is that passionate engineering doesn't always lead to commercial success.
> I respect that they’ve created their entire networking stack from the ground up. Few can claim the same. But objectively, the less eyeballs on a piece of code, the more risk. What will the company do if they’re hit by a bus?
AFAICT, they didn't write their own network stack from the ground up but based it on OpenSolaris and many of the devs were former Sun employees. That's far from "write you own network stack". Joyent started in 2004, when Linux didn't have DTrace, containers, and likely didn't compare in networking performance, so it would've made a lot of sense to run a cloud startup on an OpenSolaris base. They also shared drivers and network code with IllumOS and other OpenSolaris forks. Not exactly a bus factor of 1.
Joyent was one of the first container companies, which were based on OpenSolaris Zones, but seem to have just missed the "Docker" container wave. That and Joyent never quite hit on the same developer-oriented Docker CLI and Dockerfiles otherwise, who knows what would have happened.
I still think of Zones as superior to Linux namcespaces/conainter setup in many aspects, much less back in the early 2012-2014 when containers were just becoming "a thing". Nowadays it's mute with AWS's Firecracker and CPU support for virtualization there's essentially no performance penalty for VM's like back in 2012.
The last couple of decades mainstream OSes have been catching up with mainframe world.
Chefs that care deeply about cookware will either go one of two ways, flame out or Michelin star.
It's a hugely risky strategy but the pay off from staffing a business with people who care incredibly deeply is potentially very high as long as they're well managed. Early days Apple is probably one of the best examples of it, but Microsoft is probably the excellent counterpoint to that as well.