After Sun goes out (NewsForge) [LWN.net]

After Sun goes out (NewsForge)

Posted Oct 2, 2003 20:18 UTC (Thu) by deatrich (guest, #25) [Link] (9 responses)

Well, I read Eric's article earlier today. It's a bit doom-and-gloom, but I have to say that I really wonder about Sun. I think one of Sun's big problems is that they are still the dot in the dot-com. That is their problem. The dot-com thing sent them on a tangent, and they have never 'returned'.

Sun needs desperately to refocus. They really need to understand what is going on around them. I can only presume that some people at Sun understand, and have tried (are trying) to push the company back into the right direction. For some years now, Sun has MISunderstood the lay of the (l)intel land. This is the most disturbing part of all. If this was an x-month burp I wouldn't be concerned. However it has been years - they don't get it, they never have. What a sad thing to see a company that was innovative and fresh reduced to this catatonic state.

Just the fact that their attitude and support for solaris-intel has yo-yo-ed to the extent it has is very telling.

Here we have a company that is chalk-a-block full of unix expertise. Reving up their participation in the open-source revolution should be a cakewalk for them! But they keep dropping the ball. Would someone please wrestle their stupid spokespersons to the floor and *gag* them? They simply can't open their mouths without offending the open-source movement, and saying things that are totally ridiculous, and I mean TOTALLY.

Do they _ever_ talk to open-source people to understand why many of us would never install a solaris system (or any other commercial unix system for that matter) if we can get away with an open-source OS? I can't believe how utterly painful it is to install and patch-to-date these damn things - let alone set them up (gnu-ify) so that they are vaguely useful.

I think there is still time for them to get back on track - but not much time. It would help (in my opinion) if they would open-source java. I don't know if it is any longer possible, but it would help. If it is possible, then they should do it now. If they wait for another year, I think it will be too late. Microsoft is behind schedule, and not entirely on-track, with the .Net thingy (hey, their vaunted marketing machine hasn't succeeded here, at least not yet). Sun always had an opportunity here, but not by themselves.. never by themselves.

One thing in their favour: they have launched themselves into the (l)intel world with their (albeit strange) 'Java'* platforms (I saw a Sun presentation recently: a rather purply-gnome experience). They are selling, for example, V6xx intel- based servers for competitive prices. From an intel hardware point-of-view: other *nix vendors and intel/linux wannabees pretty much suck. I am more than a bit fed up with Dell these days; HP seems to be self-destructing (okay, I admit it, I spend _way_ too much time reading theinquirer). It would take VERY LITTLE for me to move away from Dell Linux workstations and servers. Dell's spokespersons are just about as bad as Suns. Their web sites are atrocious. Within some fuzzy price and support range I would be perfectly happy to purchase other red hat linux-installed systems.

There are opportunities here. With all that money in the bank, and with all that *nix expertise, Sun should be able to do an about-face that would set the IT industry back on their heals. Sun? Are you there? Wouldn't it be nice to buy intel-based hardware from a vendor who isn't constantly bending over backwards for Microsoft?

After Sun goes out (NewsForge)

Posted Oct 2, 2003 20:31 UTC (Thu) by smoogen (subscriber, #97) [Link] (2 responses)

Sun isnt going to change until the board and upper executives are gone. Since that isnt ever going to happen.. I have to agree with Eric's bleak painting.

On the Dell article... Dell is right, people werent buying Linux installed desktops at the rate they needed them to. Of course this may have to do with the fact that their Desktop division never wanted to sell any and had to be brought kicking and screaming to the table by the big boss. Their server division has a completely different take on Linux and have sold quite a bit.

On the other hand.. unless they were to sell at least 1 Linux desktop for every 100 windows desktops, they would lose money overall... if they only see sales of 1 Linux desktop for every 1000->10000 desktops.. they wont have to have any sort of incentive to keep selling them.

After Sun goes out (NewsForge)

Posted Oct 2, 2003 20:49 UTC (Thu) by deatrich (guest, #25) [Link] (1 responses)

I guess for Dell that desktop = optiplex.

Maybe they should just stop selling their linux workstations too then. They really annoy me. What annoying person at Dell decided that the default mouse type for a linux workstation is a 2-button mouse? And it seems to be a crap-shoot what you will ulitimately get from their online web-ordering process.

Sorry for the rant, but I have to say that purchasing linux desktops is, and has never been, an enjoyable process. There are opportunities here, really. Some of us just want to avoid the white-box experience, and don't mind paying more.

After Sun goes out (NewsForge)

Posted Oct 2, 2003 21:11 UTC (Thu) by smoogen (subscriber, #97) [Link]

I just dont think that this is where the community of do-it-yourselfers shoot themselves in the foot. I know many organizations that purchase tons of Dell Precisions with windows and then strip out and put Linux on it. The price differential isnt enough for them to have two different procurement methods.

Nit: dot in dot com,

Posted Oct 3, 2003 1:39 UTC (Fri) by kmself (guest, #11565) [Link] (5 responses)

Sun was "the dot in dot-com", literally. They used to run the root DNS nameservers. The DNS root is dot. Not that you could figure this out from the ads... Mind that Sun lost this contract to boot, a couple of years back.

Nit: dot in dot com,

Posted Oct 3, 2003 15:42 UTC (Fri) by Baylink (guest, #755) [Link] (4 responses)

Well, no.

The DNS root servers used to be run by *a lot of different people* -- look at an old root.cache file; isi.edu, PSInet, UMD, ISC, the Army, NORDUnet, etc.... none of whom seem to have been Sun.

And, anyway, the root isn't the "dot in '.com'"... it's the (almost always *implied*) dot *after* .com: "baylink.pitas.com." Truly strictly speaking, it's the *empty space after* that trailing dot, but even I'm not usually that pedantic. (UPS commercial pause) Yes I am.

Now, they're *all* run by Verisign, a topic that is an *extremely* sore sport for DNS junkies... see last weeks' piece about the .com/.net wildcard fiasco.

Well, except for the ORSC roots. Using them, of course, doesn't fix the wildcard problem -- that's in the *GTLD* root servers for com and net... but at least you're not at Verisign/ICANN's worthless behest for other issues -- and I've been using them for 2 years, in production, and haven't ever had a problem.

Nit: dot in dot com,

Posted Oct 4, 2003 9:15 UTC (Sat) by rise (guest, #5045) [Link] (2 responses)

Actually VeriSign runs the .com and .net GTLD servers, the roots are still run by a mess of independent cooperating organizations (of which VeriSign is only one). For example f.root-servers.net is run by the ISC as a distributed virtual server, M by the WIDE project in Tokyo and K by RIPE out of London and Amsterdam. See http://www.root-servers.org/ for all the gory/interesting details. Thanks, by the way, for explaining the null label - after it surfaced in conversation last night I was surprised to note that some very clueful people had no idea it existed.

Nit: dot in dot com,

Posted Oct 6, 2003 16:12 UTC (Mon) by Baylink (guest, #755) [Link] (1 responses)

I'd fallen off the domain lists (when Verisign yanked them with no
notice to avoid litigation some years back); things are less ugly than
I'd thought; thanks.

But to reply to wcooley, no, I don't think that it was ever true that all
the root servers ran on Sun hardware either... I'm fairly sure that
ISC's, in particular, does not.

Sun on the roots

Posted Oct 18, 2003 17:18 UTC (Sat) by shane (subscriber, #3335) [Link]

I can't speak about other roots, but the RIPE NCC's root server (K) ran on Sun boxes for a few months only. The fact is they were too slow to keep up with the load immediately after going on-line. Admittedly this is a failure of the organisation to properly benchmark the hardware required, but it is also not a ringing endorsement of Sun either. ;)

The K root now runs Dell PC's.

Nit: dot in dot com,

Posted Oct 6, 2003 5:22 UTC (Mon) by wcooley (guest, #1233) [Link]

He meant that the root servers were all running on Sun servers, not being run by Sun itself.

After Sun goes out (NewsForge)

Posted Oct 2, 2003 21:51 UTC (Thu) by proski (subscriber, #104) [Link] (7 responses)

Sorry for stupid question. What did Sun do? Why does ESR write that they 'crossed the line from "troubled" to "doomed" yesterday'? Did they support SCO publicly? Did they file for chapter 11? Did they fire Bill Joy? I see their stock tumbling on September 30, but I cannot find anything in the news.

After Sun goes out (NewsForge)

Posted Oct 2, 2003 22:40 UTC (Thu) by StevenCole (guest, #3068) [Link] (6 responses)

Bill Joy did leave recently and this big news on Monday resulted in stories like this.

After Sun goes out (NewsForge)

Posted Oct 3, 2003 2:57 UTC (Fri) by tjc (guest, #137) [Link] (5 responses)

From the second link:

"In a very candid and unusual letter, Milunovich urged Sun to cut 5,000 to 7,000 jobs to speed a return to profitability and recommended that it eliminate its Sparc, Mad Hatter, and Java product lines."

One has to wonder, what in the world would they sell if they eliminated Sparc and Java?

I recommend that MSFT eliminate Windows and Office. :-)

After Sun goes out (NewsForge)

Posted Oct 3, 2003 3:55 UTC (Fri) by StevenCole (guest, #3068) [Link] (4 responses)

From this Info World article:

In his letter, dated Oct. 2, Merrill Lynch's Steven Milunovich recommends Sun de-emphasize its Sparc hardware architecture and focus on the Intel x86 platform, spin off its crown jewel Java programming language, and pass on battling Microsoft on the desktop with the newly announced Java Desktop System, formerly known as "Mad Hatter."

Remember, this is Merrill Lynch talking, and if you can't trust their always trustworthy advice, whose can you trust?

After Sun goes out (NewsForge)

Posted Oct 3, 2003 12:48 UTC (Fri) by beejaybee (guest, #1581) [Link] (3 responses)

C'mon off it - AMD & Intel seem to be fighting to get their sales of 64-bit systems into double figures. Compaq killed Alpha so Sparc is just about the only realistic choice for genuine 64 bit hardware. It ain't as good as Alpha but it ain't bad either, just way too expensive & too tied to a proprietary OS.

Intel seem to be in denial that the (server) world is rapidly approaching the need for > 4 GByte address spaces. AMD simply don't seem to have the clout. In any case IA32 architecture - which is really an inefficient kludge from the original 8086 CPU, now 25 years old - can't last forever.

IMO getting out of Sparc and into Intel architecture would be the worst possible thing Sun could do i.e. shooting themselves in both feet & emptying the magazine into the space between their ears.

Sun don't have the right model for me, but they do seem to be able to shift Sparc boxes. Good luck to them, as long as they keep out of bed with SCO.

After Sun goes out (NewsForge)

Posted Oct 3, 2003 15:16 UTC (Fri) by tjc (guest, #137) [Link] (1 responses)

Intel seem to be in denial that the (server) world is rapidly approaching the need for > 4 GByte address spaces.

I'm pretty uninformed when it comes to high end hardware, but I'm vaguely aware that Intel is using a 36-bit address bus on the Xeon, which should increase the address space from 4 GB to 64 GB. Is this correct, or am I "remembering" something that never happened?

PAE mode

Posted Oct 3, 2003 16:11 UTC (Fri) by StevenCole (guest, #3068) [Link]

The 36-bit PAE mode has been around on Intel IA-32 for a long time. Here is a rather old exposition on The 64GB memory thing. Keep in mind that on 32-bit systems, processes usually only have access to 3GB on typically configured linux kernels, although an experimental 4GB/4GB split option is available on the 2.6.0-test -mm series kernels.

Interestingly, the default Microsoft split is 2GB/2GB, with 3GB/1GB being an option if you've paid more at the pump. More details are here.

64 bit processors

Posted Oct 5, 2003 9:31 UTC (Sun) by chenno (guest, #15744) [Link]

There's also IBM's G5. Now that Apple dumped Motorola, they stand a chance to stay ahead of the race.

I'm sad every time a chip architecture is dumped, but Sparc really seems to be lagging in processing power.

SUN is dead, long live SON

Posted Oct 2, 2003 21:51 UTC (Thu) by anandrajan (guest, #146) [Link]

Perhaps AMD and SUN should merge and the new company---let's call it SON as an acronym for Solaris on Opteron---should drive Solaris 64 bit on Opteron at the high end with SuSE 64 bit and 32 bit linux at the low end. These two can merge down the road at some point with Solaris bits gradually going open source and into 64 bit Opteron linux.

Anand

_______ is dead?

Posted Oct 3, 2003 0:49 UTC (Fri) by stevenj (guest, #421) [Link] (4 responses)

People periodically say that such-and-such big company (usually Apple) is "dead" and they are almost invariably proved wrong. I'm not an economist, but my informal impression is that such companies are so huge that they have to shrink a lot before they disappear...they have a lot of time to change course.

Of course, dot-coms with little or no revenue and huge paper valuations could disappear in a puff of smoke. However, in Sun's case, we're talking about a company with over $11 billion in revenue and almost $6 billion in cash reserves (according to 2003 SEC filings). They have a long way to go before they fall.

Apple was almost dead

Posted Oct 3, 2003 1:33 UTC (Fri) by funkor (guest, #15677) [Link]

It doesn't necessarily take that long for a poorly managed big company to collapse. In 1995
-1996, Apple really was in deep sh*t. It's worth noting that the company didn't really recover
until the CEO and most of the board was replaced.

If I were a Sun board member, I'd be asking some hard questions of the executive staff right
now. I agree with the poster above, there seems to be an egregious lack of oversight at Sun,
and the board probably needs to go.

_______ is dead?

Posted Oct 3, 2003 2:54 UTC (Fri) by gdt (subscriber, #6284) [Link] (1 responses)

People periodically say that such-and-such big company (usually Apple) is "dead" and they are almost invariably proved wrong.

As with DEC? Or Commodore? Or most of the Seven Dwarves?

What I find irritating about the article is that it assumes that the PC vendors have good strategies. HP's strategic direction is worrying. Dell have just shot their Linux strategy in the foot (they only support Red Hat Enterprise Linux, making their low-end servers considerably more expensive than the white box competition).

_______ is dead?

Posted Oct 3, 2003 10:34 UTC (Fri) by djabsolut (guest, #12799) [Link]

The comparison with Commodore (C=) is on the right track. C= had a good product for a while (the Am*ga), but miserably failed to innovate/develop it or place it in a coherent direction (anybody remember the CD32 debacle?). It's as if they deliberately consigned themselves into a niche and never bothered to strike out from it; C= became complacent, thinking that the niche will last forever.

Sun is doing the same - it's placing itself (whether deliberately or inadvertently) into the high-end server niche (and it is becoming a niche due the presure of Linux & Windoze utilizing the ever increasing horsepower of run-of-the-mill Intel processors). They keep on touting that Solaris on x86 is the best. What planet are they on? Why on earth would anybody want to pay them when Linux, or even the *BSD distros, are just as good or better?

The solution for Sun is a simple and an obvious one, yet they're living like they want to turn the clock back. They can keep on peddling their high-end servers, but that by itself is not enough. They will also have to focus their energies into making & selling Linux boxes and support contracts (a la Red Hat). It's called evolution: mutate or die.

_______ is dead?

Posted Oct 3, 2003 17:41 UTC (Fri) by rfunk (subscriber, #4054) [Link]

Usually the companies die by being bought out (though sometimes they call
it an even merger) and then disappearing within the new owner. Not by
disintegrating on their own.

After Sun goes out (NewsForge)

Posted Oct 3, 2003 6:10 UTC (Fri) by pglennon (guest, #649) [Link] (1 responses)

uh... don't they have like six BILLION dollars in cash? actually, I think it's 5.7 Billion, but that leads me to believe that it might be too early to call this one....

And somehow, I doubt that Sun's troubles have anything to do with SCO. That sounds like a bit of a "raymondism" ( god love him ).... sometimes you have to sift through these to get to the quality of the stuff he produced on the SCO case so far...

After Sun goes out (NewsForge)

Posted Oct 3, 2003 12:33 UTC (Fri) by pyellman (guest, #4997) [Link]

> uh... don't they have like six BILLION dollars in cash?

Yes, they do have $5 billion in cash. As such, you can assume that some proportion of the current value of their stock (current $3 or so) is directly linked to investor's valuation of that cash pile. Purely for the sake of discussion, let's say that $1 or $1.50 of the current value is linked to the cash reserves. Now, consider from an investor's point of view the implications of additional quarters like the last three at Sun; basically, the value of the cash reserves is being pissed away at frightening rate. Sun has a very large workforce, operations, and research efforts to support; at the current rate of losses, their postion is untenable, even in the short-medim term. This is a doomsday scenario, but it is also the current reality, and will produce continued declines in Sun's stock as realization spreads. I expect investors to make some kind of forceful effort soon to get Sun to articulate a new strategy, at the very least for protecting those cash reserves. With no voices inside Sun or on the board to challenge McNealy, however, such an effort is likely to fail.

Peter Yellman

After Sun goes out (NewsForge)

Posted Oct 3, 2003 10:52 UTC (Fri) by kunitz (subscriber, #3965) [Link]

Sun had made following mistakes:

1) Sun didn't listen to the customer

- no virtual servers
- new hardware required Solaris 8, forcing us to keep our leasing
contracts for older hardware

3) Screwed customers during dotcom time

- for instance: a single sale of 50 Sun Enterprise 450 (quadratic
4-way-boxes with a lot of air in it) equppied with one processor
- price ratio between Kingston RAM and Sun RAM 1:8 (Notify that
RAM from Sun has been manufactured by Kingston); of course
support contracts forced you to buy Sun RAM

4) Didn't provide SPARC workstations priced against PCs

This way they guaranteed, that Sparc hardware vanished from
almost all developer desktops around the world.

5) Unbundled C/C++-Compilers from the OS and didn't support GCC

This way they minimized the number of binaries tied to the SPARC
platform. Yeah, if you want to play the lock-in game, you have to
play it right.

6) the Java thing

It always strikes me, how they are able to combine the
disadvantages of proprietary software development (long feedback
cycles, great expenses for QA and bug tracking) with the
disadvantages of open source (no revenue stream) in Java.

Java is quite good, compared to C# (more mature class library, faster
after the slow startup). However it will be never an option on the
desktop, until they cache Jitted code. IBM's SWT works around that
problem, but doesn't solve it. Compare startup times of make and ant.

After Sun goes out (NewsForge)

Posted Oct 3, 2003 16:12 UTC (Fri) by freeio (guest, #9622) [Link]

Honestly, one of the most counter-productive moves Sun made was to make sure that their Solaris and app software licenses for any box expire when the box changes hands, unless it changes hands through Sun's program. Yes, there is a free binary program for educators and developers, but it covers only Solaris, not the apps.

The result, of course, is that either GNU/Linux or one of the BSDs becomes the easiest thing to put on the box. This is called losing mind share, and losing market share, and it is being done quite intentionally. I assume that this is being done to keep up the perceived value of Solaris, but it is a loser all the way around.