panick21_

Except Amazon wasn't really doing they had loses because they invested in scaling.

Viking I am talking about was around 1992.

Of course by mid-2000s that is different.

That's a bad comparison. Falcon 9 wasn't developed to be reusable. And it it didn't actually take that long. Also Starship has already proven that Starship booster can land. And Starship itself has cleanly landed on the ocean.

There is still a question about turnaround time, but just comparing the state now to early Falcon 9 is not accurate.

Now the point of economic sustainability was always putting Starlink into orbit. That's something people in the space community have said from the beginning.

And both the booster and the Starship itself have proven that they can get back to ground mostly intact. So its not like they are only working on 'getting to orbit'. And clearly it is leading something desirable if the have proven that they can basically land the second and first stage, something nobody else has ever achieved.

Question how fast they can make the turn around are still there along with other question. But once you can repeatably launch you can improve much faster as they did with Falcon 9. And you can do more and more tests in shorter time, accelerating development.

I honestly think its a incredibly silly thing to bring up over and over. But I guess that's what it like to be in public. I did hear the podcast and checked the argument.

You can find the interview here:

https://www.youtube.com/watch?v=-vDunup8v5M&t=2940s

I'm not totally convinced by the reasoning. Would love to hear some former Sun people give an opinion on his leadership and other perspective.

> Yeah, and it was too late to edit my comment. My bad. But yes. The SPARC folks had some great ideas in the T line, but it was a) way too late, b) didn't have access to latest node size, so it wasn't going to be great-performing, c) performance sucked.

Dropping performance a little bit for Multicore was reasonable for servers, and there were some good ideas there but in practice it was far to extreme to be practical.

StorageTek was quite expensive and sold mostly very large archiving solutions. Not sure that was really the right direction. Such a huge acquisition cost a lot, not just money.

But you are right that this was not totally unreasonable. Still its pretty aggressive move for the company in the position of Sun.

> Sun needed: internal disruption.

After the disaster that was Viking. They might have gone 32bit x86, 64bit Alpha. DEC was desprate for Alpha buyers and would have made them a good deal. But they double down on going all in on SPARC and UltraSPARC confirmed them in this being the right solution.

I think it worked out for the best. That said if we had Linus implementing LVMS instead of Linux history might not be that different.

I think one serious issue was that they took way long to realize portability.

They had rough group implement MICA OS that was supposed to be partially compatible. But the dropped MICA and PRISM and instead adopted MIPS and Ultrix. But then they never bothered to port VMS onto MIPS and their workstation line.

Then they mismanaged their whole VAX line and massive lost out to the improving Unix offerings, partly for performance and partly for 'openness' reason.

I think if they were smart they would either have gone 'all-in' on MICA and make it the future of both Unix and VMS. Or simply support both Unix and VMS on all of their platform. They were big enough that this wasn't some incredibly complex task.

They ported VMS to Alpha, but that was a 64bit machine that most people didn't need. Meaning they lost in the lower spec server spaces where Sun and HP were selling tons of Unix servers.

Had they ported VMS to MIPS and also had sold lower spec server with all these features, VMS might not have become so niche.

Personally I really believe that their complete mishandling of RISC, despite having arguably the best chip design team in the industry to be the reason for much of their failure.

Suns designs were not excellent. In fact their history with CPU making is pretty damn terrible. The made the wrong choice almost the whole time.

They got it right initially, making Solaris with gate-array and that was a big success. Small company got RISC chip out fast and it beat what was out-there.

Then they went all in on CMOS-Bipolar chip, that was massively late, massively expensive. The completely lost the plot and were not competitive in the early 90s. Thankfully they had the software buy in from costumers, and did some good stuff otherwise (they also did the disastrous Solaris transition) but they overcame all that.

Then they did a bit better with UltraSPARC in 1995, they were still not top of the line, getting easily crushed by Alpha and others. But it was 64 bit and that helped them long term. Nowhere near 'excellent' for the time period. More like, good enough. Then they rolled on from there, but were never anywhere close to Alpha and others until around 2000.

They made all the mistakes you can make in the 90s, going heavy into Gallium Arsenide that was a disaster. Then they went heavy on VLIW technology in another disaster.

Then they completely lost the plot in the early 2000s, and eventually pretty much gave up completely and switched to marketing T1 chips and 'threw-put computing' because they couldn't produce competitive chips on single thread performance.

In fact they did switch from Texas Instruments to TSMC in that time period. You can read about that in 'Life Under the Sun: My 20-Year Journey at Sun Microsystems'. Don't take everything in that book at face value, its very much putting a positive spin on things, but it tells some of those stories.

I mean Olsen was kind of right. Most VAX sold didn't run Unix and VMS and VAX were incredibly successful. Unix at that time was successful on workstations more then servers.

And by the time Unix became successful it was as complex arguably VMS was at the time.

DEC didn't fail because they supported VMS, in fact VMS was great costumer lock in for them and its still made money for HP despite being developed in the late 70s.

And in terms of clustering, Unix has never matched VMS.

DEC didn't fail because of VMS. Their big failure was not to make a fast RISC workstation that ran VMS early enough. And then in then in the late 80s not switching to RISC fast enough with VMS.

So I think he had a point, most serious banks and insurances did run VMS and paid lots of money for it for the next 20 years. VMS was not the reason DEC ran into issues. Their complete management of first personal systems, then workstations, then mainframes, then RISC servers is what destroyed them.

You can read 'LIFE UNDER THE SUN: My 20-Year Journey at Sun Microsystems' if you want to know what the leader of SPARC said at the time. The book does a lot to sweet talk some of the issues, but also admits to them.

Well he became leader when it was clear that they were deep in the shit. Then he was ask if SPARC should just be killed of. He said know but said he would need 4-5 years to turn it around. And I would argue he didn't even if they made some waves with Sparc T1 chip.

Ultimately I would argue anybody reasonable would go back to conversation and say, we need to kill SPARC and switch to x86 as fast as possible.

Cantrill of course was wrong in saying what he said, but he was actually right about about the argument they were having.

Oral History of Mark Himelstein goes into that:

https://www.youtube.com/watch?v=-vDunup8v5M

He said he killed x86 but wanted to it only temporary, as some sort of political move around Itanium.

I think it was not a technology that should have been designed for people actually doing series work. But for things like kiosk, self service terminals, and stuff like that.

And yes of course, a rack of commodity servers should how powered the server side, as this was a task that could be scaled vertically quite well.

The also didn't use the cheapest possible Sun Ray terminals, they used their internal developed embedded Sparc chip, not mass produced ARM chips.

As long as it is local its fine, I mean X also doesn't have much security, its not that big a problem.

Basically no real use-case actually are pure integer like that. Basically it performed badly on pretty much all workloads that anybody cared about.

> - Cancelling -if briefly- Solaris on x86 in 2002. This killed Solaris in the minds of many who didn't want to be locked into Sun for SPARC.

There is an interesting interview with they guy responsible. He basically even said it was a mistake and he knew it, and he internally signaled 'don't worry its just temporary' but there was a cooperate need to signal support for SPARC. Seemed to be internally political.

I think he didn't realize how externally this would look for the viability of Solaris on x86.

> - Closing Sun PS (professional services). Bad bad move, possibly the worst of them.

Can you give some context. I am interested in Sun history and I don't know that one.

> - UltraSPARC was more than a decade too late to make up for SPARC falling way behind x86_64. Sun needed to give up on SPARC, but again, vendor lock-in sounds sweet but turns out to be poison.

What do you mean, UltraSPARC was in 1995. A decade earlier was pre commercial RISC.

(Edit: I see you mean Sparc T1, but those chips were trash, they were not to late. They were good marketing above good product)

UltraSPARC was one of the few actually successful developments. Unlike pretty much every other CPU project at Sun starting in the 90s.

> - The MySQL purchase. WTF, this was horrible and stupid.

Also StorageTek.

True but Apple is a consumer company that makes consumer products. Sun lost all interest in that already by the early 1990s. They were happy to just be standard unix and didn't do anything to be a real consumer company. They talked a lot about competing with Microsoft but didn't actually.

They talked about buying Apple, and they could have, but the CEO basically admitted they had no clue what so ever on what to do with Apple.

They had no vision or concept of what being a consumer company would even be like. Even in the 80s they were focused on engineers and costumers like that, never end costumers.

No its not 'obviously inefficient'. And just asserting competition is 'meaningless' isn't actually an argument.

Neither food, nor shelter, nor healthcare are natural monopolies. As evidence by the fact that non of them are remotely monopolistic in places where there are no restrictions. Monopolies in these sectors are created and enforced by the state. Unless you are using a completely different definition from all economists.

Ironically major shops sell stables such as rice below price or at basically 0 margin, just to get people into the store where they might buy something of higher margin.

The subsidies are not 'enormous' vs the total value of all food consumption or production.

There is a huge difference also between subsidizing demand with private supply and providing supply. Exactly the debate that is relevant in healthcare as well. A system can do both, or neither or a combination of the two.

Subsidies account for a very small % of either production or consumption of food.

Not sure why you are making an argument about food subsidizes to me, I have not argued in favor of them.

> More than half of all calories produced in the US are not even for human consumption.

Most of those go threw animals and are then consumed by humans. There is a lost factor there but saying its not for human consumption misses the point I would say. For Chicken, Pork and fish the lost factor is pretty small.

Its amazing to me how Americans (and others) act as if there are only 2 things in the world the US (private) system and public system that apparently everybody else on the planet has.

The US system is neither fully private nor are those other systems fully public. Its a false binary people use so they don't have to actually engage in the details of how different systems operate.

Nobody is arguing that the US has a good or 'the right' system. And nobody should argue that the British NHS (or systems like it) is some amazing magical place that gets everything right.

> When will the American people wake up and realize their private healthcare model only benefits the rich that are able to afford the care.

And yet one of the countries wit the lowest spending % of GDP on healthcare with very good outcomes has a system that has a lot of private aspects.

> Healthcare in America is not accessible, extremely expensive, and inelastic.

Did I somehow make the argument that the US has a good health care system? I must have missed that.

> Healthcare is the only service that is life and death, where the money paid by the patient is completely inelastic.

Most healthcare spending isn't actually 'life and death' either. And even if it in the medium term is 'life and death' only a very small amount of health care is an emergency where you can not potentially consider different options.

Just as most time when you are going to the store its not 'life and death'. But still if don't do anything eventually you will starve.

> The fact remains clear: The number one cause of bankruptcy is medical debt.

My goal would be to achieve the best overall healthcare outcomes at reasonable cost. That's at least the first stab at a definition, I'm sure we could come up with a better criteria.

If you increase payroll taxes as much as a fully public system would required, you would just have more people default on their houses and cars instead. Because there is no world, no matter how many billionaires you disown, that you can implement a universal system without massive increases of income taxes.

Again I'm not even arguing that a universal system is bad or wrong thing to aim for. The argument that I actually made was that 'some people make profit' isn't actually a good argument against any system.

> Don't act like public healthcare, expanding access to primary care physicians, and reducing medical administrative bloat is some economic degeneration headed for the sickle and hammer. I'm tired of pretending Ayn Rand style libertarianism capitalism is the perfect solution that has worked for everything including healthcare.

You are literally just ranting about shit I have never argued. I guess I must have trigger something.

If you want to make the argument that anything that is essential should be publicly owned you can make that argument. But I would suggest if you want to make that argument 'nobody should make profit' shouldn't be the relevant argument. The argument should be 'this is how we can solve X problem efficiently'.

Not enough doctors in the pipeline. The government gave doctor associations control over the pipeline and basically even encouraged reduction, because it increases wages. They don't want to many doctors. Lawyers do the same potentially even better.

I'm sure burnouts happen but the fundamental problem is the pipeline.

Of course people need to take some responsibility, but reality is most people live in a society and do as the society does. You can minimally change people with some financial intensives but the effect of that is pretty limited and its hard to implement.

You have to create a healthy system, that takes social engineering and government cooperating on a general plan. And in the US such a thing is basically not possible, even if they wanted to do it.

Ideally your transportation, agricultural, educational policies should all work together to produce healthy population. But this simply isn't the case.

But it will not 'reduce pressure' in a practical way, because such changes purely private or public take decades to work themselves threw the population.

How the US does scheduling is kind of insane. Doctors like pilots (and truck drivers in Europe) should have clear rules about how long they work with rest periods and so on. Giving people 24h shifts is nonsensical and insane.

By that logic all food production and distribution should not be privatized because is should be for the public good rather then the profit of the few.

There are are arguments to be made about private or public but that logic makes no sense.

Some system that have many private aspects outperform other public systems and the other way around. And even in a public system many people will profit as many of the companies that are your suppliers are still for profit.

The engines have proven repeatability and some have re-flown on a orbital rocket. By the standard of any company outside SpaceX, the Raptor has achieved remarkable things. Its at minimum the second most reusable engine in the world.

It has not yet proven to be as reliable reusable as Merlin, but the test programs they have put Raptor threw are considerable.

Raptor has successfully landed a rocket, and has successfully landed even an upper stage. Making it the only engine ever to achieve both.

If at this point you call that 'premature' then you must be an Asian parent in terms of your exception.

Tory was just wrong. Yes TVC wasn't there but that's because that's not inherent to the design, some Raptors are attached with TVC some without. But everything else was there.

And if you see Raptor integrated on Starship you can see this clearly.

Vertical integration is one thing SpaceX does but its not unique and magical as many other space companies have tried the same without achieving results similar.

I will not attempt to break all the reasons for SpaceX success in this post, but its certainty more complex then just vertical integration. And vertical integration isn't absolute, in some cases they reversed their vertical integration again.

Team building by Musk, consistent having engineers work on hard project after hard project, willing to invest in a next generation even if the existing system works in the current market context (canalizing yourself), believing in increasing scale overtime (only when planning to build 100s of engine iteration makes sense) are just some of the relevant parts of SpaceX success as an organization.

People claiming Musk has nothing to with this or is just some irrelevant figurehead money guy are idiots. At the same time, the brilliance of the team he build ultimately is far more important then his own suggestion about ideas with engines.

In case of the engine its actually Tom Mueller who basically off hand suggested to him that to reach his goals, he would have to design something nobody had ever done on a large engine, and was only used for smaller engines (phase shut off). Musk jumped on that and imitatively said that this is what they should do. Mueller said he wasn't sure if it could be done, but since they knew of no other way to achieve the goals, the team went off and eventually achieved it.

Mueller himself said he thought going after this was a mistake and a less aggressive simplification should have been done, but said he proved himself wrong and it was a massively important innovation.

This shows some important aspects of Musk. Not that he is some genius that magically design an amazing engine. Rather that he listened to his engineers and gave them the finances and confidence that they could achieve outlandish targets, and continued to support them even if they couldn't hit the target instantly. We see this again with Raptor, they had a target in mind and followed every line of design to achieve something close to as good as a 2 fuel engine is going to get.

One can argue that this is just blind optimism and then throwing money at the engineers, and that is certainty partly a fair critic of this approach. It did not work with self driving in the same way at Tesla. But the difference is that with evolution of Merlin and Raptor he listen to engineers who proposed challenging ideas that were still iterative and he supported them.

I have followed Space for a long time. Musk isn't some magical unique genius, but he certainty isn't do nothing idiot some of his haters want us to believe.

SpaceX success deserve series study and SpaceX without Musk is basically really understandable as an institution.