Hacker News Reader

Comment by jph | Hacker News Reader

Sun and open source was a fascinating collision of management, trying to balance the sales-driven groups who aimed to sell enterprise server hardware, and the software-driven groups who wanted the freedom to run anywhere on anything.

Then-CEO Scott McNealy had exactly the right vision of computing as a utility, with future-forward projects including Jini networking, JavaSpaces tuples, wearble Java ring, Sun Ray thin clients, Netscape/Mozilla alliance, NetDynamics app server, mobile-first web pages, heterogeneity using LAMP stacks and Intel chips, and more. Sun had all the right pieces and all the right engineers to create an offering akin to the AWS launch.

https://wikipedia.org/wiki/Jini

https://wikipedia.org/wiki/Tuple_space#JavaSpaces

https://wikipedia.org/wiki/Sun_Ray

https://wikipedia.org/wiki/Netscape

https://wikipedia.org/wiki/NetDynamics_Application_Server

View on HN

Sun hardware was very popular through the 90's into the early 2000's, before Linux was considered "stable." After that, a Sparc running Solaris wasn't competitive with an early 2000's era Intel box running Linux. Around the same time, the dot-com crash flooded the market with barely used, almost new hardware. Anyone who needed Sparcs had them... for cheap. After 2003 or so, I never saw any newer companies go with Sun.

Still, I loved Sun hardware. I had a Sparc 5 at home for a long time.

I was a heavy sunos/solaris user and honestly, when they came out with the sun USB keyboard with all the sun-isms, I switched to solaris x86 at home and didn't notice any difference (except I could afford fast hardware)

> Sun had all the right pieces and all the right engineers to create an offering akin to the AWS launch.

Yeah:

* https://en.wikipedia.org/wiki/Sun_Cloud

We had a project for utility computer with sun. We would rent a data center and they’d turn machines on for us on request.

5 months before EC2. I'm not sure what early EC2 was used for, but it looks like Sun Cloud was mostly batch jobs?

Yeah. You'd submit a script. Not designed for what made EC2 popular.

there's a ton of good tech here but tbh adoptability was difficult. tuples spaces, jini, &c are still bleeding edge tech, that can possibly do amazing things, but finding out how to marshal those capabilities has been non-straightforward, knowing to what ends to put the tech, for a decade+.

thin clients make hella sense, are still something I lust for linux being able to do well (they're not great many-user graphical desktop hosts now).

sun had great systems but the buy in was usually high. utility compute (aws) eschews that.

Check out an HP t520 or t620 thin client on eBay. They’re cheap, have free management software and run Linux.

They ship with VMWare, Citrix, RDP, terminals, browsers, etc and work great.

the management software is good to have but heavens i would stay away from such antiquated junk as this. get 1L business pc cheap, with an i5-6500t, for around as much money, and get 10x better a system. radically more modern.

but my problem was not with finding client hardware. my problem is with running the things the thing clients attach to.

linux just doesn't really have it's stuff together to make use of video cards well. there's some very very expensive, rather aged video cards that can do multi-desktop. i think august 2014's amd s9150 might be more of the more competent offerings, if i remember. as i said: aged. and using bad old proprietary software. linux just does not have it's stuff together to make good multi-user use hardware resources, for powering thing client type systems.

Oh gotcha.

That’s never going to happen, because nobody has an incentive to fund the development beyond the basics. User experience is too much of a secret sauce and is expensive to do. Look at how much $ Apple spent to put a high quality UI on Unix... nobody will give that away as it doesn’t deliver benefit to them.

There's some good systems like Apache Guacamole and x2go that help a lot. But afaik everything is software based, which is probably really fine in many cases, but I crave more. It proves though that people are willing to fund the development, that the open source ethos of going further by going together is already happening.

With Wayland in particular, I think perhaps more is possible. That we might be able to get multiple users sharing video hardware. Gnome for example just landed headless support[1], which is physical-output-free desktops. Currently intended more for debugging, but Linux has been emerging better & better primitives for accessing GPUs for years, slowly extracting the capabilities of the X monolith into sensible system level constructs, and hosting a motherloading ton of desktops for folks is one of the great possible upsides of all this long long labor.

[1] https://gitlab.gnome.org/GNOME/mutter/-/merge_requests/1698

> thin clients make hella sense, are still something I lust for linux being able to do well

It is a complicated area. Sun got somewhat successful with SunRays once they had virtualization on the servers and could run Windows. Quickly after iPhone and iPad came which are not as thin, but offer many of the same benefits with more power on the device.

The remaining niche for Thin Clients is very small.

As someone without a lot of knowledge about that time in computing's history... so what happened? Is there a good place for me to go read up on this story?

Thanks!

Ironically, some people blame the failure of Sun on Jonathan Schwartz being too open and not generating enough revenue.

Ironically, some people blame the failure of Sun on Jonathan Schwartz being too open and not generating enough revenue.

What killed Sun was classic innovators dilemma, they could have pivoted into making highly reliable x86 kit running Solaris and existing SPARC customers would have lapped it up, but they held onto to trying to sell SPARC for too long and by then their SPARC customers realised that they could make do with commodity x86 kit, just buy twice as much of it and shrug when it broke.

The bigger picture was that Intel got "good enough" and while SPARC performance was still better, it started to lose badly on price/performance and the software industry adapted to that by architecting and writing code suited to the Intel model of many cheaper machines.

Reply on news.ycombinator.com