Of Sun Ray laptops, MIPS and getting root on them

In Sun’s ideal world, a user would run programs on a central server (a Sun, of course), having their session follow their smart card seamlessly from terminal to terminal along with any other shared resources they might require. While Sun produced the JavaOS-based JavaStation in 1996 — ironically based on Oracle’s Network Computer concept — it used relatively expensive hardware, being essentially a miniaturized SPARCstation 4. Instead, the new proof of concept for a cheaper, more connected world was the 1997 NetWorkTerminal “NeWT” — one wonders if that abbreviation was a coincidence — based on Sun’s MicroSPARC IIep CPU, and that prototype in turn evolved into the first Sun Ray thin client in 1999, codenamed Corona.

Setting up a number of Sun thin clients – both fixed and laptop models – running off a Sun Ultra 45 workstation, with smartcards and all, is basically my retro computing end game. I have always been deeply fascinated by “the network is the computer”, and while there’s quite a few other thin client platforms, it’s the Sun one that feels like the real, original concept. I can’t find a reasonably priced Ultra 45 anywhere – feel free to contact me if you have one on offer – but Sun thin clients are a dime a dozen on eBay.

In any event, it will come as no surprise that I love the linked article.

Follow me on Mastodon @[email protected]

Back when I was in college (’01 through ’06), the computer lab in the College of Engineering was fully fitter with about 200 SunRay clients, along with the smart cards. It was an interesting concept. The most incredible part was we logged into the SunRay station only to be forwarded to a Windows desktop. This was around the time that I got into Linux, so the whole thing sounded ludicrous to me… Using the Unix infrastructure to serve Windows desktops. Additionally, these thin clients didn’t have the capabilities that most people expect from a computer. Most importantly, there’s no way pass files to and from the system. For example, you cannot plug a USB drive to the thin client. I guess if they had set up a server to upload and download files, it would’ve worked, but that’s not what they did.

At the end of the day, the concept, at least as applied in the lab, was doomed because the system delivered less capabilities for way more money.

teco.sb,

At the end of the day, the concept, at least as applied in the lab, was doomed because the system delivered less capabilities for way more money.

Haha, yeah sounds a bit silly, but I’ve seen similar configurations using windows terminal services before. Even the web browser was served by terminal services. I think the idea behind getting everyone on virtual desktops was to eliminate the IT burden of managing local desktop hardware & software. I guess they achieved this, but meh. IIRC windows terminal services was able to forward USB storage so it was fairly transparent in that respect, but performance-wise this centralized approach can leave a lot to be desired compared to the resources of having local desktops running in parallel.

I think sun managed to use resources more efficiently with solaris where profiles migrate smoothly over the network but all the software runs locally. Active directory sort of did this but I always remember there being exceptions and software still had to be installed on local computers so it didn’t work as transparently as solaris.

Did your school previously use these to connect to Solaris?

Where I was, which was the first half of that time range, we had thin clients that could connect either to Solaris (X) or Windows (ICA.) IMHO it worked incredibly well – it meant students would frequently use Unix as their day-to-day desktop environment, in an era where people didn’t do that on home machines, so it was a learning experience. Despite the “network is the computer” tagline, a lot of the benefit was because students all connected to the _same_ Solaris machine, which meant being able to see each other’s processes, send messages, share files, email, look at somebody’s web site, etc. Windows of the era was more limited (it was 32 bit only) which meant having a larger number of servers, and they did a better job of hiding users from each other. The only thing Windows did really well was printing, which “just worked” on Windows but was a pain from Solaris. USB sticks at that time weren’t so common (I saw my first one towards the end of this time range), and the Internet was beginning to prove itself for communication (eg. you should ssh to the machine from home, scp files, email things to yourself, etc, etc.)

I don’t know if the same result would happen today though. At the time what was interesting was seeing how much more interaction students had by being on the same device, which allowed for more communication options. Today we have more distributed communication options (eg. social networks) so the benefit would be less pronounced.

malxau,

Despite the “network is the computer” tagline, a lot of the benefit was because students all connected to the _same_ Solaris machine, which meant being able to see each other’s processes, send messages, share files, email, look at somebody’s web site, etc.

I’m quite confused by that. Why would any of those benefits be dependent on connecting to the “same” solaris machine? With the exception of “see each other’s” processes (and I’m perplexed by why this is important to you?), we could do everything else you mentioned without having to connect to the same machine. Maybe your school set things up differently, but at my school everybody was connected through a network file system providing access to user/group files regardless of the lab and computer you connect to. If you always needed to connect to the same computer all the time you’d loose most of the networking benefit’s of solaris IMHO.

Windows of the era was more limited (it was 32 bit only) which meant having a larger number of servers,

Intel introduced physical address extensions in 1995. While user processes remained 32 bit, the sum of process memory could exceed 32bit address space, this wikipedia article says intel supported 64GB of ram.

https://en.wikipedia.org/wiki/Physical_Address_Extension

So I wouldn’t think 32bit windows requires more physical servers on account of 32bit process address space. But given that windows sessions are predominantly graphical, it seems likely every windows user would have needed more memory at the server even though PAE enabled servers to have more than 4GB ram.

and they did a better job of hiding users from each other.

Yeah, it may just be a holdover from older more innocent times, but I feel a lot of old unix conventions used horrible security practices: showing user processes, overly granular POSIX permissions, /etc/passwd being world readable. It seems like security got bolted on an afterthought.

The only thing Windows did really well was printing, which “just worked” on Windows but was a pain from Solaris.

Are you speaking as a user or administrator? Is there any chance the complexity had to do with the thin client? We printed stuff all the time, it just worked (and we used them a lot because our CS printers were free whereas printers in other departments and library had to be paid for). However I was just a user, I am not aware of whether solaris printing was difficult to set up for the admins.

USB sticks at that time weren’t so common (I saw my first one towards the end of this time range), and the Internet was beginning to prove itself for communication (eg. you should ssh to the machine from home, scp files, email things to yourself, etc, etc.)

I don’t ever remember ever using a usb stick on those unix systems. We transferred everything online and every computer had a public IP that we could connect via X/SSH/FTP/etc. You could even write your own daemons and connect to it using a public IP. I’m honestly not sure what they do today. I guess these days students may have to connect through a VPN.

My file transferring practices have regressed though on account of android making it such a pain to network files. I really despise android’s inability to mount network shares directly without having to manually copy files using 3rd party apps.

I don’t know if the same result would happen today though. At the time what was interesting was seeing how much more interaction students had by being on the same device, which allowed for more communication options. Today we have more distributed communication options (eg. social networks) so the benefit would be less pronounced.

Students could interact by way of shared groups. email, chat clients, etc. Using the same physical machine never played into it for us, so I’m very confused by what you mean.

Why would any of those benefits be dependent on connecting to the “same” solaris machine?

There’s a lot of old Unix facilities that were designed for multiple users on one machine. See w/who, talk/ytalk, utmp, and we had some homegrown things for this too. Honestly though, I’ve tried to have this conversation with many people who’ve come from the PC world without success – people always tell me it’s possible (which is true) but having lived in both worlds I observed the lower friction of being able to see what people are doing and start interacting. At some point it becomes an exercise in trying to explain a rainbow to somebody colorblind – some things you just have to experience.

Strictly speaking though, we did have NFS mounted home directories, and maybe had NIS authentication (I don’t remember that.)

Intel introduced physical address extensions in 1995.

Windows didn’t support this until Windows 2000, and our servers were NT 4. (I think there was an NT 4 Enterprise Edition that did support PAE, but these devices were Terminal Servers, and there was no Enterprise Terminal Server.) The Solaris machines were basically a SPARC equivalent of PAE, using a 64 bit kernel with a 32 bit userland.

Are you speaking as a user or administrator? Is there any chance the complexity had to do with the thin client?

As a user, and nothing to do with the thin client. Windows has a defined printing model where all programs could observe and print to the same queues. At the time, printing on Solaris meant setting up an lpr command line, but that meant doing it on a per-program basis. So in one case you click a printer and go, in the other case you remember the command or your job isn’t going anywhere. I’m sure it would have been possible for an Administrator to pre-bake this into all common programs via system configuration, but being per-program it’s a larger task and in practice just didn’t happen.

malxau,

There’s a lot of old Unix facilities that were designed for multiple users on one machine. See w/who, talk/ytalk, utmp, and we had some homegrown things for this too. Honestly though, I’ve tried to have this conversation with many people who’ve come from the PC world without success – people always tell me it’s possible (which is true) but having lived in both worlds I observed the lower friction of being able to see what people are doing and start interacting.

I find it fascinating how different our experiences were on the same platform. While I am familiar with them. I can’t remember a single time using them to collaborate with other users at school. The only time I remember using talk was with someone on a BBS in completely unrelated environment. Instead at school we used irc and other clients. Maybe this is because our lab situation was different from yours. We didn’t use thin clients to connect to solaris, we were logged in directly at the local console and naturally students would find themselves on different machines every day.

Windows didn’t support this until Windows 2000, and our servers were NT 4. (I think there was an NT 4 Enterprise Edition that did support PAE, but these devices were Terminal Servers, and there was no Enterprise Terminal Server.) The Solaris machines were basically a SPARC equivalent of PAE, using a 64 bit kernel with a 32 bit userland.

Ok, I didn’t realize microsoft didn’t implement PAE sooner.

I’m sure it would have been possible for an Administrator to pre-bake this into all common programs via system configuration, but being per-program it’s a larger task and in practice just didn’t happen.

I confess I’m not familiar with solaris admin, but is it possible there were default settings & print commands that worked for us? If not I guess they might have preconfigured it. I used the printers many times from text editors and web browsers and don’t remember having any issues.

Did your school previously use these to connect to Solaris?

I cannot say, for sure. But I vaguely remember the lab being “recently refurbished” into this new scenario. The interesting thing, I found, was that for our Finite Element Analysis course, we had to log into a SunOS 4 server in order to use the I-DEAS software. It was not possible to access this server from the computer lab with SunRay thin clients. Instead, we had to go to another computer lab that had standard computers which dual-booted a Red Hat version and Windows. The Red Hat install was configured to start only an X server and connect to the SunOS server with I-DEAS installed.It was a real mess.

I can’t find a reasonably priced Ultra 45 anywhere – feel free to contact me if you have one on offer – but Sun thin clients are a dime a dozen on eBay.

Oh, ffs Thom, for the third time or so:

The price that Sun Ultra 45 machines are going for on eBay currently (around $1500) is reasonable once the original price and storage costs for a 15-year-old desktop are taken into account. Real estate is not free or cheap, so anyone who held on to such a machine for 15 years doesn’t want to have done it for free. And anyone who wanted to get rid of their Sun Ultra 45 for clearance prices (so it won’t take up valuable physical space) has done it years ago.

So, stop demanding that people serve you for free (that is, that they keep a machine stored in decent condition for you to come and buy it when you realise it exists). If I stored anything of value for 15 years, even a literal paperweight (let’s assume said paperweight is valuable because it’s designed by Philippe Starck or something and there aren’t too many around), I would like to be compensated for the service. Otherwise, I would have thrown it away years ago so it doesn’t occupy valuable physical space (if I didn’t need it for personal use).

kurkosdr,

Oh, ffs Thom, for the third time or so:

The price that Sun Ultra 45 machines are going for on eBay currently (around $1500) is reasonable once the original price and storage costs for a 15-year-old desktop are taken into account.

It would be a valid point to justify the price by talking about supply and demand. However since you focused on original price and storage costs instead, I actually think that’s wrong. All that matters is what the market will bare today in terms of supply and demand, not how much was paid originally or how much it costs to store it.

I’m thinking specifically of “computer reset”, a computer store that horded a massive amount of archaic computer equipment in a seemingly endless warehouse.https://www.youtube.com/watch?v=rvM82T3C2Ik

They had more stock than even museums and collectors knew what to do with. In the end, since their supply >> demand, it turned into a buyers market and they encouraged visitors take as many machines as they could because otherwise they’d have to get shredded & recycled. The point being, storage costs are incidental, the actual value for something several decades later is determined by the supply / demand curves over time, which may or may not be profitable for whoever stores it.

All that matters is what the market will bare today in terms of supply and demand

Agree, and note that when it comes to older systems, supply is constantly contracting.

AFAICT these things follow a bathtub curve – initially things are expensive where there’s more demand than supply; after a few years people are throwing them out since they’re obsolete and there’s more supply than demand; but once supply has contracted until it meets demand, supply can’t grow again, so prices rise until the nostalgia factor is gone. We don’t know how long that takes with computer equipment since the field is too young.

Even more strangely, the nostalgia crowd disproportionately want the old, high-end stuff, not the old, low-end stuff. The high end stuff wasn’t high volume from the outset, so demand can exceed supply dramatically. I’d be shocked if an Ultra 45 is cheap any time soon, but there’s plenty of nice Sparc machines on eBay for people who want them.

malxau,

Agree, and note that when it comes to older systems, supply is constantly contracting.…Even more strangely, the nostalgia crowd disproportionately want the old, high-end stuff, not the old, low-end stuff.

I agree with all of your points.Regarding nostalgia, some brands have a lot more of it than others.Another example is original 50y/o star wars lunch boxes going for ~$1500 on ebay, whereas road runner, dr suess, hanna barbera, and similar lunch boxes from the same period are going for ~$100. It’s quite a large discrepancy in supply / demand for products that likely came from the same factory and probably cost the same originally. It’s easy to see the trends in hindsight but predicting trends 50 years into the future is hard because of social uncertainties and other unknown variables.

It would be a valid point to justify the price by talking about supply and demand. However since you focused on original price and storage costs instead, I actually think that’s wrong. All that matters is what the market will bare today in terms of supply and demand, not how much was paid originally or how much it costs to store it.

I thought of saying something about supply and demand, but everyone understands it and my post was too long already, so I didn’t. But the two concepts aren’t mutually exclusive. if you’ve kept a desktop computer in storage for 15 years in decent condition and nobody wants it, you have still spent the time and money (real estate is not free) to do that, you’ve just taken a loss.

But when supply and demand are on your side, why on earth should you offer that service for free to the people who think your time and real estate have no value? The price those eBay sellers are asking for a Sun Ultra 45 is reasonable once the storage service is taken into account, even if it’s supply and demand causing it. If the price was something like 10000 bucks, that would be high (but still, you don’t get to define what is “reasonable”).

kurkosdr,

I thought of saying something about supply and demand, but everyone understands it and my post was too long already, so I didn’t. But the two concepts aren’t mutually exclusive. if you’ve kept a desktop computer in storage for 15 years in decent condition and nobody wants it, you have still spent the time and money (real estate is not free) to do that, you’ve just taken a loss.

Well, to use a common expression, a broken clock is still right twice a day. The fact that you’ve sunk money into buying and storing something over time might correlate with a higher price, but then it might not because it’s not the causal factor in determining what you can sell it for. I think the post about lunch boxes highlights this quite clearly. Two people can “invest” in physically identical products that only differ in branding. But despite both people incurring the same sunk costs, their profits (or losses) can be entirely different because they’re facing different supply and demand when they go to sell.

But when supply and demand are on your side, why on earth should you offer that service for free to the people who think your time and real estate have no value?

It’s not up to me or even you to say whether someone’s time or services have value, in a market with adequate liquidity, free market principals will decide it. Real estate is no exception. Sellers naturally want to believe their time and effort in improving & remodeling adds value and feel they deserve a ROI above and beyond their investment costs. In reality though investments don’t automatically translate to profits if it’s not inline with market conditions. It comes back to the same theme, if market conditions are in your favor, then great, go do the investment. But if they’re not in your favor, then the truth is that committing your time, effort, and money might not actually be valuable regardless that you think it should be.

The price those eBay sellers are asking for a Sun Ultra 45 is reasonable once the storage service is taken into account, even if it’s supply and demand causing it.

I still believe the market will be indifferent to what prices you feel are “reasonable”. You probably think I’m being too ridged on this point, but just as Thom’s opinion that things should be cheaper bugs you, your incidental rational for why they’re more expensive bugs me. Oh well, I think we may be beating a dead horse, can we agree to move on? Till next time, cheers!

Jaysus, who pissed on your cereal this morning?It was just a simple tongue in cheek comment.LOL

You will get the same feeling when you hear the same canned complaint for the 10th time.

It’s a silly tongue in cheek request. Unless you are sitting on a pile of boxed Microsoft bob boxes hoping that they are worth something in 30 years, I have no idea why this would bother you. Ha ha.

In any case, hopefully Thom learns that there is this thing called qemu where he can run solaris for Sparc on an emulator on a modern PC faster than a Sun Ultra 45 ever did. Or just run solaris for x86 on a VM and get even better performance for the craptastic experience that is solaris on modern times.

These old systems are just capacitor ticking bomb anyways. So unless Thom is comfortable with a soldering iron, I don’t think he realizes what he’s asking for anyway. ha ha.