Good article, I liked what he wrote about the GNU tools and package management. At my last try a few weeks ago I failed with OpneSolaris again because of they ancient GCC version to compile the necessary software elements, SUN's cc failed on that code with a different problem while it's just fine on Linux or a OSX.
I wonder if/when it's ever going to occur to them that it's easier to work on improving linux than it is to fight windows, mac AND linux by ... becoming more like linux.
I especially loved this point - "Will buyers accept a certain amount of lock-in, or will they insist on zero barriers to exit?" Sun acts like it still owns the market.
I don't know, google seems easy to get out of. Not that you'd ever want to leave, of course. Mail/calendar/docs/rss can all be exported. Any artificial restrictions I'm missing? App engine yeah, but it's barely up yet.
This might have been better as an email to one's executive team, rather than a public blog post.
I'm not convinced about being a deployment platform. Sun's repeatedly been eaten from below by "cheaper" alternatives of arguably lower quality (whether Wintel, Linux, etc.). I'm not sure I buy that focusing on incremental quality/performance improvements is going to be a winner now either.
Sun is the General Motors of the computer industry. They used to do tons of stuff for everyone because their products and expertise were unrivaled. As the demand for big machines picked up in the 90s, they did well and sold a lot of stuff to a rapidly expanding market. But now that we've all figured out that smaller, cheaper machines are just as good, nobody can quite imagine how they'll regain their past glory.
And that's the problem with recovery plans like this one: they're full of fine ideas, and you could imagine the company picking a few and doing well at them. But could they even match the efficiency and hard work of their scrappy Asian competitors? They're gonna need some major inspiration to flip around a multi-year trajectory of slow failure. The bus just doesn't have that kind of turning radius, I'm afraid.
agreed that sun needs to make some bold moves (besides changing the stock ticker from SUNW to JAVA) in order to get out of the mess they are in.
the changes that the writer suggests are bold and do require a new team and a lot of capital. looking at sun's balance sheet <http://finance.google.com/finance?q=java>, it doesn't look like they will be able to do much of anything. i think sun is sitting around and barely making do because they just really want someone else to come in and take over. they've been slowly cutting jobs for years in order to reduce OP EX.
If I was running Sun I'd seriously look at splitting the company in two. The software side is shackled by the fact that they need to move in a direction that is going to erode revenue from the hardware side. As long as they are in the same company you will have a tug of war between two sides pulling two different ways and nobody going anywhere.
I agree with the software folks trying to become the best deployment platform for web software. There's a window of opportunity while the JVM is ahead of other VMs where Sun could pull this off if they can produce a super stable, scalable, easy to deploy runtime that runs everything from PHP to Python, Ruby and Groovy and make it a snap to deploy and manage. The catch is that I don't think there's any chance they will achieve this if they try and tie it to either Solaris or Sun hardware. That ship has sailed - hosting these days is either linux or windows for all but the highest top tier. They need to completely get over it and aggressively push themselves on linux.
come to think of it, sun may be the only company to have the entire stack of technologies in house... chips, servers, os, VM, programming language, development environment.. cloud computing seems an obvious direction.. perhaps they are doing that already?
But why should it be? All those environments you mention can run equally well on any operating system. If Solaris on SPARC can be made price/performance superior to Linux on Intel, then that's what they should do.
It's hard to see exactly what Linux gets Sun apart from "me too" points. They don't want to be Dell.
> If Solaris on SPARC can be made price/performance superior to Linux on Intel, then that's what they should do.
No - this is where I think the worlds collide. You're basically saying they should commoditize their own hardware to the point where its at parity with Intel. I'm saying they need to do the opposite - they need to increase (or at least maintain) their margins and focus on selling it to the people who really need it - enterprises and top end institutions that value the incredible robustness and scalability you can only get with that kind of hardware. I think trying to do both ends of the spectrum on the same platform is just going to ensure they do neither effectively.
Well, not necessarily. If a pair of Sun Exxx's in a cluster was cheaper, faster and more reliable than a rack or two's worth of 1U Dells then it would be a no-brainer.
Right now, for most applications, AMD64 beats SPARC. I would love to run Solaris on AMD in production. I can't tho' because Solaris x86 doesn't get tier-1 support from a couple of vendors I rely on. The choices are SPARC/Solaris or Intel/Redhat, and despite the fact that Red Hat is a complete pain in the arse, I can't justify SPARC/Solaris on grounds of better manageability and I can't do AMD/Solaris either. That is where Sun could make some quick wins.
Tim does a good job at highlighting some key tech held by Sun. But there is no clear thesis that even if there is significant adoption of these technologies that Sun makes money.
I do hope some of these technologies keep moving forward. I do not see a great way for Sun to profit from them.
One of Solaris rock-solid advantage over Linux was the guaranteed stable ABI. Except Sun shot themselves in the foot by introducing Java which made it irrelevant...
I just installed NetBeans 6.5 "ruby edition" after reading this post and just had to come back here to explain why Tim's thesis of "we have many great complementing products to create a coherent whole" (my words, not Tim's) is mostly an anchor around Sun's neck.
I download NetBeans 6.5 (ruby only version) and the installer tells me it needs to use 145 MB for the install. I tell it ok and it asks for my root password and I give it and away it goes. The install gleefully tells me at some point it is installing "GlassFish v3 Prelude". WTF!!! Stop doing things like this!!!
About a year or so ago, I seriously took into consideration adopting Solaris 10. I read the basic install docs and tried to get a clean and basic install of Solaris. No such luck. I wanted a server OS and the default install gives me not 1 but 3 desktop solutions, multiple Sun products like a directory and mail server and yes, 3 versions of Java setup in a "great" multi-java install method. This was done to enable the multiple other stuff to work as sun's various products it was jamming down my throat had different Java requirements.
So what would I need to get a "slimmed down Solaris install"? I would need to read through a thousand pages of docs and write install scripts.
So what did I do instead? I switched to ubuntu server. Why is ubuntu server great? Because it does not have other products it is trying to "up-sell" me.
Sun thinks that because it is giving you some free tech that it spent lots of money creating that they have the user's permission to shove more stuff on the user. This is a big problem. If I want GlassFish, I'll install it. If I want any form of X-windows or Java on my server, I'll ask for it.
I installed it yesterday (on Windows) and there's a big fat "Install Glassfish v3 Prelude" checkbox on one of the first pages. Unticking this should work for you. I'll give you the benefit of the doubt 'cos perhaps the *nix installer is very different
I just did an install of Solaris 10u6 yesterday and it was dead simple. During the installation, the installer asks you what software group you want to install. Select the "Core" software group or the "minimal networking" software group.
I agree the documentation is confusing. That is because (1) Sun documents everything so there is a lot of it to weed through, (2) most installs of Solaris are done over networks using install scripts to ensure every installation is identical, (3) Solaris can do things like install a new OS on a remote system that doesn't have any OS on it, and the documentation for those things drowns out the smaller, simpler documentation for the simpler installation methods.
I would guess their major problem is lack of mindshare. All other problems more or less come from that, including what appears to be a lack of focus.
They need to borrow a page from Apple's book and build a SPARC-based "Sun mini" that's cool, desirable and cheap. It should also be the obvious choice for a desktop platform for anyone who wants do develop for Java, Solaris or whatever Sun thinks is strategic.
Getting CMT chips in the hands of maverick developers is vital if they want people to develop for them. Parallel stuff is seen as cool and sexy and their CMT SPARCs are built for it. I would buy a desktop Sun to "test the water", but I wouldn't do the same with a rackmount server. If they have rackmount servers going for USD 2700, they sure can do desktops for about USD 1000. If they can make Ruby or Python do great threading with them, all the better as I have no real desire to program in Java.
Back in the OpenWindows or CDE days, Solaris wasn't a terribly nice desktop environment but today, with Gnome, it is every bit as good as, say, Ubuntu. The right people should start seeing the Sun logo more and the huge userland Unix-like OSs have is key.
And, while they are at it, the JAVA ticker was a stupid idea. Sun is, and should be perceived as much more than that. SUNW was not a brilliant one (as workstations were no longer relevant)
31 comments
[ 5.0 ms ] story [ 72.3 ms ] threadI especially loved this point - "Will buyers accept a certain amount of lock-in, or will they insist on zero barriers to exit?" Sun acts like it still owns the market.
And so do all the other cloud providers.
I'm not convinced about being a deployment platform. Sun's repeatedly been eaten from below by "cheaper" alternatives of arguably lower quality (whether Wintel, Linux, etc.). I'm not sure I buy that focusing on incremental quality/performance improvements is going to be a winner now either.
And that's the problem with recovery plans like this one: they're full of fine ideas, and you could imagine the company picking a few and doing well at them. But could they even match the efficiency and hard work of their scrappy Asian competitors? They're gonna need some major inspiration to flip around a multi-year trajectory of slow failure. The bus just doesn't have that kind of turning radius, I'm afraid.
It's worse than that. All the kit that Sun sold in the 90s is still perfectly usable. Why buy new even if you want Sun?
Wasn't Patrick Naughton asked what Sun should do when he resigned?
Seems to be the new black - winning the mind share, losing the battle.
http://www.wired.com/wired/archive/3.12/java.saga_pr.html
Except that Sun is (as far as I can tell) failing at everything they are spreading their peanut butter on.
the changes that the writer suggests are bold and do require a new team and a lot of capital. looking at sun's balance sheet <http://finance.google.com/finance?q=java>, it doesn't look like they will be able to do much of anything. i think sun is sitting around and barely making do because they just really want someone else to come in and take over. they've been slowly cutting jobs for years in order to reduce OP EX.
I agree with the software folks trying to become the best deployment platform for web software. There's a window of opportunity while the JVM is ahead of other VMs where Sun could pull this off if they can produce a super stable, scalable, easy to deploy runtime that runs everything from PHP to Python, Ruby and Groovy and make it a snap to deploy and manage. The catch is that I don't think there's any chance they will achieve this if they try and tie it to either Solaris or Sun hardware. That ship has sailed - hosting these days is either linux or windows for all but the highest top tier. They need to completely get over it and aggressively push themselves on linux.
Unless they move into the cloud domain - in which case, their expertise in both areas (software/hardware) will be complimentary and useful.
Not really - the one piece of the puzzle that Sun doesn't really have is services (aside from support).
It's hard to see exactly what Linux gets Sun apart from "me too" points. They don't want to be Dell.
No - this is where I think the worlds collide. You're basically saying they should commoditize their own hardware to the point where its at parity with Intel. I'm saying they need to do the opposite - they need to increase (or at least maintain) their margins and focus on selling it to the people who really need it - enterprises and top end institutions that value the incredible robustness and scalability you can only get with that kind of hardware. I think trying to do both ends of the spectrum on the same platform is just going to ensure they do neither effectively.
Right now, for most applications, AMD64 beats SPARC. I would love to run Solaris on AMD in production. I can't tho' because Solaris x86 doesn't get tier-1 support from a couple of vendors I rely on. The choices are SPARC/Solaris or Intel/Redhat, and despite the fact that Red Hat is a complete pain in the arse, I can't justify SPARC/Solaris on grounds of better manageability and I can't do AMD/Solaris either. That is where Sun could make some quick wins.
I do hope some of these technologies keep moving forward. I do not see a great way for Sun to profit from them.
I download NetBeans 6.5 (ruby only version) and the installer tells me it needs to use 145 MB for the install. I tell it ok and it asks for my root password and I give it and away it goes. The install gleefully tells me at some point it is installing "GlassFish v3 Prelude". WTF!!! Stop doing things like this!!!
About a year or so ago, I seriously took into consideration adopting Solaris 10. I read the basic install docs and tried to get a clean and basic install of Solaris. No such luck. I wanted a server OS and the default install gives me not 1 but 3 desktop solutions, multiple Sun products like a directory and mail server and yes, 3 versions of Java setup in a "great" multi-java install method. This was done to enable the multiple other stuff to work as sun's various products it was jamming down my throat had different Java requirements. So what would I need to get a "slimmed down Solaris install"? I would need to read through a thousand pages of docs and write install scripts. So what did I do instead? I switched to ubuntu server. Why is ubuntu server great? Because it does not have other products it is trying to "up-sell" me.
Sun thinks that because it is giving you some free tech that it spent lots of money creating that they have the user's permission to shove more stuff on the user. This is a big problem. If I want GlassFish, I'll install it. If I want any form of X-windows or Java on my server, I'll ask for it.
I agree the documentation is confusing. That is because (1) Sun documents everything so there is a lot of it to weed through, (2) most installs of Solaris are done over networks using install scripts to ensure every installation is identical, (3) Solaris can do things like install a new OS on a remote system that doesn't have any OS on it, and the documentation for those things drowns out the smaller, simpler documentation for the simpler installation methods.
They need to borrow a page from Apple's book and build a SPARC-based "Sun mini" that's cool, desirable and cheap. It should also be the obvious choice for a desktop platform for anyone who wants do develop for Java, Solaris or whatever Sun thinks is strategic.
Getting CMT chips in the hands of maverick developers is vital if they want people to develop for them. Parallel stuff is seen as cool and sexy and their CMT SPARCs are built for it. I would buy a desktop Sun to "test the water", but I wouldn't do the same with a rackmount server. If they have rackmount servers going for USD 2700, they sure can do desktops for about USD 1000. If they can make Ruby or Python do great threading with them, all the better as I have no real desire to program in Java.
Back in the OpenWindows or CDE days, Solaris wasn't a terribly nice desktop environment but today, with Gnome, it is every bit as good as, say, Ubuntu. The right people should start seeing the Sun logo more and the huge userland Unix-like OSs have is key.
And, while they are at it, the JAVA ticker was a stupid idea. Sun is, and should be perceived as much more than that. SUNW was not a brilliant one (as workstations were no longer relevant)