As far as I could tell perhaps the #1 Sun died was fantastically inept management, most especially in sales:
Sun just wouldn't talk to you directly if you weren't buying mainframe levels ($$$) of hardware. At some point they established a site where you could punch in your credit card, but if you needed more you were at the mercy of their "VARs" and resellers, who in far too many cases also wouldn't give you the time of day (note this was CPU agnostic, it was true for their x86 hardware according to the stories below and for SPARC hardware in my personal experience).
A little quality time with a search engine will find stories from all sorts of startups upset that they were forced to buy Dell since Dell actually wanted to sell stuff to them in any quantity (at the same time HP's sales function was also messed up in different ways).
So Sun missed out on all the startups that became really big; by the time these companies were buying in mainframe quantities Sun was just a bad memory.
One thing that I noticed happening later was companies dropping Sun because their hardware kept changing. In one case they dropped a great discrete Intel Ethernet chip in favor of the ones on the nVidia chipset, in another, they kept changing the management interface. (A bit like one of the things that hurt Compaq, you could buy many copies of the same model, yet adjacent serial numbers might have totally different motherboards.)
It's no accident that Sun's sales organization seems to be where Oracle is making the biggest quickest changes (hiring thousands of salesmen as I recall).
ADDED THOUGHT: I don't think the thesis in the linked article flies. As I see it, it posits that if Sun had done a (much) better job with open source, the business gained from that could have covered for their failures elsewhere.
Better, I say, to have fixed both sets of failures.
What you're describing is a classic case of channel conflict. The same problem largely killed off several CAD software vendors back in the day. It was also one of the key reasons why Novell's NetWare business failed after being highly profitable for years.
Isn't this more like "channel incompetence" than channel conflict? My understanding of channel conflict is that it happens when manufacturers disintermediate their channel partners by selling directly to customers (which is exactly what it seems Oracle will be doing now).
What Sun did was to fail to police their channel partners and confirm that they would actually sell intermediate quantities of equipment; I know that throughout the '90s it was extremely difficult for me to buy single Sun SPARC systems (workstations and servers), and as noted in my original message too many startups found it impossible to buy six figures worth of systems.
I assume their partners understood that Sun handled the really big purchases (maybe they got a piece of the action???) and that the only conflict was when Sun set up their web storefront for the small purchases a customer could put on a credit card.
It is hilarious to me that the Linux community continues to retell the "Have you ever kissed a girl?" chestnut after 14 years, and still fails to get the joke.
If you go back and read the original Usenet (not email) drama (http://www.cryptnet.net/mirrors/texts/kissedagirl.html), it is Miller who sounds like an adolescent. From describing himself as "chief architect of the SparcLinux port," to the laughable wishful thinking of "Linux is light weight, Solaris is a pig," to the non-sequitur cheapshot of "Sun Quentin" (what is that supposed to mean? that Sun engineers are held against their will? that they're felons? what?), the entire post drips of the naive self-regard that, sadly, marked Linux's kernel community at the time.
I was regularly using both systems in 1996. Solaris worked. Linux did not. Miller was asking to be taken down several notches, and Cantrill did so in the fewest possible syllables.
"Sun Quentin is the informal nickname of the offices of Sun Microsystems. Close proximity to the San Francisco bay and the architecture of the buildings are reminiscent of San Quentin, a California state prison." [from urbandictionary.com -- and personal experience]
Man reading that reminded me of why I disliked Linux and all of the "hackers" in that space at that time. The adolescent smugness was so thick back then. That email really took me back. OS zealotry seems to have died down since then, that or I have just gotten too old to care or notice.
It's still a religious debate. I just talked about it with a professor in the field and he refuses to have a conversation about OS with other similar folks due to the lack of rationality in the debate.
They may have missed the joke, but the problem is that adolescent obsessive compulsion to prove oneself does actually accumulate to certain kinds of advantages in the long term. Cantrill really was underestimating what he was up against.
I think a confusing hardware lineup, confusing messages and overall confusing strategy have a whole lot more to do with its merger than any compulsion to control code.
The difference between Sun systems and everything else out there in the late 1980’s (mostly Novell Netware or Microsoft LAN Manager networks) is that Sun networks just worked, all the time.
Ugh, not quite. I installed and configured an early version of DECNet-PC (ultimately renamed PathWorks) on a 4-node Vax 8650 cluster in late 1986 that networked three Epson PC's over ethernet to a common virtual DOS drive mapped on an RA81. It was solid as a rock, behaved itself well, and easily shared files for WordPerfect, Lotus 1-2-3 and dBase.
An early Netware (2.11?) network had been setup three floors above us to show the power of 'client-server' computing. It crashed at least 5 times daily when it finally would run and its unreliability got two of my co-workers fired because they couldn't make Novell's empty promises work.
Lan Manager? Late 1980's? Are you kidding?
Sun started by unseating DEC in the scientific desktop market with their "workstations," not their network.
Just look at the nodes on maps of the early ARPANet if you don't believe me. PDP's, VAX's and mainframes.
When Scott McNealey announced that the "network was the system" he was taking direct aim at DEC whose installed systems already made up the world's largest collection of networked computers and pc's, were clustered to 16 nodes and more, offered full load balancing, were tied to gigs and gigs of storage using high-speed teflon interconnects and routinely ran 200 days and more without a restart.
Don't get me wrong, Unix/NFS was nice, but it needed a lot more work to compete with DEC, VAX/VMS and Ultrix. What prevented DEC from eating Sun's lunch was it's insistence on centralized computing using terminals. Had they fully understood what they had in DECNet-PC back in 1986 there never would have been a need for Novell, or Banyan Vines, or WFW 3.11 or any of the other wannabe's and Sun's "networked desktop workstation" would have been a much more difficult sell.
Here's the spec for a VMS filename circa 1985:
NODE "accountname password"::device:[directory.subdirectory]filename.type;ver
Meaning: Use the text editor "edt" to open version "4" of the file named "mylog.txt" that's located in the "configs" directory under the root directory named "sysfiles" on drive "0" attached to disk unit "a" on the node named "orion". Log me in using the "system" account with password "somepwd."
From anywhere in the world as long as it's on the network.
Everytime I configure Samba I think about those Epson PC's.
Yet Apple survives and thrives. Sun wasn't really a hardware company or a software company. That was the problem, it seemed they really didn't know what they wanted to be, so they tried to be everything and failed.
Apple counterpoint is right on the "money". Apple was having quite a lot of problems before wunderkind returned.
This is obviously hindsight, but imagine:
It is early 90s, you are the company that goes by the motto "the network is the computer", and you have this incredibly hot tech called Java that can run on any tier. A bit later and your JME is on billions of mobile devices all over the world.
A Steve Jobs type of leader would have been able to leverage that amazing potential. And the transformation would have been as decisive as Apple's.
Sun was never able to turn JME into a viable application platform. The standard library was too limited and didn't allow much access to interesting mobile device peripherals (GPS, camera, microphone, motion sensor, etc). So to write any kind of significant application required separate versions for each cell phone model and carrier using their proprietary API extensions, security system, and provisioning model. This fragmentation prevented the emergence of a developer ecosystem.
JME had a lot of potential at the beginning but Sun just couldn't execute or get the device vendors and cellular providers to cooperate on common standards that would have benefited them all. The iPhone and Blackberry have now made JME largely irrelevant.
Yes, Sun's inability to execute in consumer markets appears in the failure of applets against flash. And it would also prevent them from creating something like an iPhone.
It seems that just about all current success stories in computing are in consumer markets: search, social websites, smartphones.
It's true that Oracle is doing well and they aren't in consumer markets - but their market is not showing rapid growth; on the contrary, they are actively consolidating it.
And I guess the business lesson here is to learn to spot dissonance in your business strategy.
Sun engaged the consumer market with propaganda. That was the extent of their engagement. (Remember all those Sun ads during the dotcom bubble?)
But in reality, Sun was in reality only engaging geeks. Sun has never failed as a company catering to geeks.
Somebody in that company should have noted that their desire (Sun for the masses) and their actions (Sun for techies) were not congruent.
Proof is in the fact that in the case where the general public has little input - the backend -- Sun was (and remains) spectacularly successful. The mistakes made here are far less specific to Sun, and very much related to the dissonance between being in business to make money and being enlightened hippies who are making substantial contributions to the industry, for free.
Sun just wouldn't talk to you directly if you weren't buying mainframe levels ($$$) of hardware. At some point they established a site where you could punch in your credit card, but if you needed more you were at the mercy of their "VARs" and resellers, who in far too many cases also wouldn't give you the time of day (note this was CPU agnostic, it was true for their x86 hardware according to the stories below and for SPARC hardware in my personal experience).
A little quality time with a search engine will find stories from all sorts of startups upset that they were forced to buy Dell since Dell actually wanted to sell stuff to them in any quantity (at the same time HP's sales function was also messed up in different ways).
So Sun missed out on all the startups that became really big; by the time these companies were buying in mainframe quantities Sun was just a bad memory.
One thing that I noticed happening later was companies dropping Sun because their hardware kept changing. In one case they dropped a great discrete Intel Ethernet chip in favor of the ones on the nVidia chipset, in another, they kept changing the management interface. (A bit like one of the things that hurt Compaq, you could buy many copies of the same model, yet adjacent serial numbers might have totally different motherboards.)
It's no accident that Sun's sales organization seems to be where Oracle is making the biggest quickest changes (hiring thousands of salesmen as I recall).
ADDED THOUGHT: I don't think the thesis in the linked article flies. As I see it, it posits that if Sun had done a (much) better job with open source, the business gained from that could have covered for their failures elsewhere.
Better, I say, to have fixed both sets of failures.