If it comes to a war? The US has over 50,000 troops over there! There's no way they're going to bring the boys home without giving them something to shoot at first.
The US has *NOT* done a good job of dealing with bin Laden. He's still alive, al Qaeda is still active, anti-US sentiment is at an all-time high in the middle east. What better way to distract people from the failure of dealing with al Qaeda than a war they're guaranteed to win with someone who the US public knows as a Bad Guy? Hussein has thumbed his nose at the US so many times that he has a callous on it, and yet he hasn't really done anything in GWB's regime to justify a war. That alone is probable cause for a war.
The whole "weapons of mass destruction" argument is such a sick joke that it gives me hives. Tell me again which is the ONLY country in the history of the human race to use nuclear weapons against another country? Tell me which country has the largest stockpile of weapons of mass destruction? Oh that's right, it's the holier-than-thou US.
North Korea is violating UN resolutions left, right, and centre; but the US is still happily supporting them. The US is promising to tear up the 1972(?) international nuclear non-proliferation treaty and also start developing chemical/biological weapon programs, but they don't see anything wrong with THEMSELVES doing it. No, it's only when an arrogant Enemy Dictator(tm) with huge oil reserves (another important point) gives them half an excuse to start a guaranteed-to-win war which will distract the population from how ineffective they've been elsewhere that they MUST go to war for the good of the world.
How's that?
Time to retire the shuttle
on
More on Columbia
·
· Score: 2, Interesting
OK, we can't retire the space shuttle today. Nor tomorrow. But its time is drawing near...
Consider first that the shuttle was a massive compromise versus the original proposed designs. If the budget had been infinite, we would have had a better shuttle. If the budgeteers had had more foresight, we might (probably) have had a better shuttle. The shuttle we have now is a big series of compromises that limit its usefulness and safety.
Now consider that the shuttle program has been around since 1981. That's more than half of the time that's passwd since man first walked on the moon! It still seems shiny to some of us (myself included), because it was the only newsmaking bit of space exploration in our youth. However, it's old. It's an old (and limited) design, and we have learned a lot of what to do (or not) on the next go around. It's time to climb the next step of astronautical evolution.
So let's keep them in top shape, fly them as necessary (mostly as ferries to the ISS), while putting as much money as possible into a next-generation space vehicle.
I don't really understand why you would say that Solaris is hard to administer. The documentation is fantastic--head and shoulders above most other manufacturers out there. Solaris is an elegant system V OS, with a handful of BSD tools retained (in addition to the Sv ones) For those who want 'em.
Admintool gone? Thank god! That was unquestionably the WORST thing that Sun ever did. The Management console is better by far, although as you point out, almost unusably slow thanks to Java. Sigh.
Moving to Linux tools would be a HUGE step backwards. XFree86 config easier? Xfree86 is the single worst Unix/Linux related tool of any sort, anywhere, anyhow, that has ever been created in my opinion! I have spent over a hundred hours (!!!!!) of my time trying (often unsuccessfully) to get Xfree86 working on different boxes. It is ugly, it is horrible, it is buggy. It is BAD!!! Sun X? It just works. Turn on the computer and go.
vim and GNU grep? No point. There's nothing added in vim other than eye candy, bugs, and poor documentation. GNU grep has a few nice features, but nothing that's much better than egrep, which comes standard with Solaris, HP-UX, and AIX.
Linux has a few simple problems that you're trying to drive Solaris into embracing here. Linux suffers EXTREME and CONSTANT versionitis. The stupid dependencies and constant upgrades are silly and unnecessary. This is exacerbated by the multiple distros of Linux out there, which aren't even particularly compatible with each other, let alone with de facto Unix standards. To pick on one tool, xinetd is a disastrous mess. Any tool that replaces something that's been in existence for 20+ years, and yet doesn't maintain backwards compatibility is entirely unacceptable. Then we get to the documentation, or lack thereof. There are no doc standards, no consistent format, no guarantee of version-concurrency. The number of times I've seen in a Linux man page, "this document is not kept up to date" is just sick. The info docs aren't much better, and the how-tos are worse. Informal docs written and maintained (or not) by any random volunteer who can type is not how you write documentation.
The TCO of Linux is currently lower, because the only people who run Linux servers are knowledgable Linux-heads. If a Linux solution was purchased by a company for a professional, controlled, change-managed, production environment, then they'd see their TCO SKYROCKET vs. Solaris, or HP-UX, or AIX, or even IRIX.
On the Linux report card, there's a comment from the teacher: "Doesn't play well with others, suffers from inflated self-worth, disrespectful of playmates." Dumping a good solution for a hackers OS because YOU understand it better is a bad idea. A company spending lots of money needs more than just a robust OS (which Linux, mostly, is): They need a consistent, stable (from a development/update point of view), coherent, compatible solution. Turning all of those solutions into Linux is a bad idea.
OK, cvs sucks in the particular instance of serving up relatively static files. There should be a requirement that everyone who sets up an inappropriate cvs server (where ftp or http would be a better idea) should have to write "versions are not files" on a blackboard, by hand, in chalk, 1000 times. I keep seeing sites that have final, stable, production releases of program 'x,' and the only way to get it is cvs. Dumbness.
Now for actual version concurrency/control, cvs rocks and rocks hard!
1) He already stated this fact. 40% I/O throughput increase. (actually quite a large variability, but it's a usable number)
2) Read the subject line! He said, "THE AVERAGE POWER USER..." Now I read that as meaning the home Linux geek/developer who likes messing with the guts of their system. Companies use RAID0 all the time, or more often RAID1+0. RAID5 is equally common, implemented in hardware. This is not what he's talking about. This is not the target of his comment.
Sigh. Sorry to rant, but every follow up to this article has neglected this point.
I'm not trying to elevate him on a pedestal here, but I just don't want to see him vilified. He screwed up. Big deal. He flipped out on ONE comment out of several. Big deal. He had the decency to apologise. Not a huge deal there either. None of this is a big deal. He seems like a decent guy. That's all.
On the other hand, I'd like to see someone squeeze an apology out of John Romero.:-)
He admitted that they screwed up. (or fucked up, as the case may be.) He lost it when pivx when public. Then he apologised for losing it, and admitted that pivx was entirely in the right.
This is about as much news as the bug itself. Not much.
Bah. Don't bother with the smelly little troll. Let him crawl back into his biometrically-secured bunker, and wait for the rest of the world to forget about him.
I only have one issue with your post. I'm not sure the US has had all that much goodwill since the late-1950s. Even so, George W. certainly is burning it as fast as he can.
"They aren't, HP was first. You can have x86 blades and PA-RISC blades in the same chassis."
Are you sure about this? The only HP blades I can find are the proliants, which means Compaq, which means intel. I've also been told repeatedly that while both HP and IBM COULD have had non-x96 blades, they don't yet.
I'd love to see it. And your idea of a generic blade frame is something I've thought about too. What a nice idea that would be, but I don't see it happening.
Actually, they announced it a few months ago. Now they've finally given it some substance, although it's hard to follow.
N1 is a low-administration, self-managing environment. It's a farm of servers and infrastructure that can dynamically reallocate themselves as needed for different loads, and do so intelligently. If your accounting server is getting hammered at month end, but so is your after-hours game server (hey, it's possible!:-) then the computers themselves will know to reallocate more resources to the accounting, as it's more mission critical than gaming. Similarly, it will allow for metered semi-automatic allocation of resources. Imagine being a CIO getting a page and an email one night: "Your webserver is running at 95% capacity, and serving 'x' pages per second. For $20,000, you can double your capacity. Please reply with "yes" or "no" in the subject line, to enable this feature or not as you want." Neat plan, eh? It's something that Sun talks about at any rate. It's feasible, with some work.
DISCLAIMER: Since most of this is extrapolation from what I heard, I can't imagine that anything here violates my NDA, but in the off chance that it does, it's a lucky guess.:-)
Two weeks ago, I saw a private announcement of this technology that they talked about today. (signed the NDA and everything!:-) At that time, the blade servers they announced led me to think of two things:
1) This heterogenous blade system is INCREDIBLY cool!!! You can have Intel blades, Sparc blades, encryption blades, network caching/acceleration blades, etc. etc. etc. all in one frame. The fact that Sun (late as they are) is the first out of the blocks with this idea is remarkable.
2) I WANT CC-NUMA ON BLADE SERVERS! With all the mysterious hype Sun has given N1 (learning from MS, anyone?), I really hope that this is part of their upcoming plans. One OS instance across "n" computers, where n is a variable. Time will tell, I guess.
"Code is GOOD if it provides users with the features they need."
Hmm. I'm not sure I quite agree with this. I think we're using the word 'code' to refer to two different things.
A PRODUCT (application, tool, whatever) is good if it provides users with the features they need. CODE is good if it's well written. Code is well written if it's clean, logical, consistent, efficient, and extensible. Implied in there is that it's easy to debug as well. (which is actually a result of good code)
The old Unix programs lacked features because they weren't possible, realistic, and/or desired. When graphics hardware didn't exist, there was no need for a GUI-based editor. The fact that those applications are now obsolete doesn't make the code any weaker in the slightest! Sometimes the natural lifespan of a program comes to an end.
Think about this as a tech manual. A first edition of the Unix System Administrator's guide isn't all that useful anymore. It talks about how to build a kernel on OSes that don't exist anymore, and it's such specialised knowledge that it's not particularly transferable to modern Unices. It's obsolete. It's not relevant. The only way to 'fix' it is rewrite it from the ground up.
However, the writing style is as good now as it was then. Nemeth et al raised the bar for Unix manuals, by writing an excellent, well organised, understandable, readable, and entertaining book. None of those qualities are changed by obsolescence.
As for sendmail, it's effectively the same answer. No matter how it's written, sendmail is not the right tool for every mail job! There are very good reasons for writing smaller, lighter, easier mail MTU. There are also reasons for keeping sendmail. None of this is remotely relevant to the coide quality of any of them. If a sledgehammer is so good at its job, why would anyone waste their time inventing a finishing hammer? Or a ball-pein? They're DIFFERENT TOOLS!
"GNU versions of fileutils, for instance, corrected a lot of these limitations, at the cost of uglified source code"
Again, I disagree. the GNU tools certainly did fix a lot of limitations in the original tools (and those limitations in some cases are the result of bad code I'll admit, or at least poor choices for code), but the uglified source is not a necessary cost. There's no reason that they can't be written as elegantly as the original ones, or possibly better. From what I've seen of the current version of BIND, it's still clean well written code. There is nothing inherent about modern, 'complex,' massive computing that forces code to be ugly. Only laziness and poor practices on the part of individual programmers causes that. On a massive project, the only way to prevent that sort of bad code is to rigorously screen the programmers, and rigorously screen the code. (Note: Rigorously here means a formalised mandatory code review--having the code available for people to read and fix isn't sufficient.) Unfortunately, OSS by its very nature lacks in both of these areas. I think that was, in a roundabout way, the point that Ken was trying to make.
This is a neat theory. It's not how business works in computing.
Why buy Sun? Why would ANYONE buy Sun from the equation above? My equation had a 'take to the purchasing manager' result, and that is SOLID. "I don't care" (assuming you mean about what OS is run on it) isn't something that makes business decisions. "I don't care" doesn't sell computers. Flexibility of OSes on Sparc is COMPLETELY irrelevant. In fact, nobody really cares what Unix variant is on a box when they're buying it. They know that when they buy Sparc, they're buying into Solaris, and that combination should be unbreakable. If they weaken that link, THEN their sales will suffer.
And x86 just isn't the same market. It never has been, and it never will be. Nothing that Sun does on the x86 side will substantially affect their Sparc sales. The Intel stuff is neat and perhaps profitable, like the SunRays, but it's not an integral part of their business, and it won't sway anyone towards or away from buying a Sparc.
1) People are lazy. 2) People feel they owe their employer nothing, and have no loyalty. (perhaps partly justified) 3) The modern work ethic seems to be show up in the morning, go home in the late afternoon, and make sure your work all gets done. Whether you do that at work or at home, and when you show up or go home (within reason) seems to be less and less concrete.
If I'm given a three hour job at 4:00 that MUST be done today, and am not getting overtime for it, then be damed if I make sure my coffee breaks don't go over 15 minutes. Alternatively, if I'm on a punch clock and my workday is over at 5:00 then I'm going to work my ass off until that very moment.
I think ultimately the workplace has become more relaxed, which gives slackers and whiners more room to complain without getting their lazy asses fired. The rest of us will just enjoy having more freedom in how we get our work done. (and work more effectively for it)
"Maybe at some point state and Federal regulators will realize that the Bells are the problem, not the solution."
OK, I don't know enough about the phone situation in the US to comment on the subject, but is this line REALLY necessary? I mean it's flat out incitement and misdirection--personal opinion masquerading as factual content.
You're missing out on one of Sun's biggest points: reliability.
On the Sparc platform size, Sun builds high-end, high-reliability, high-predictability boxes. They want an OS that works (a) very well with the hardware, and (b) with the same reliability features. If they're going to promote porting of other OSes onto their platform, they'll do it on their own terms, and with their own requirements, and that's not very straightforward.
It's smartest, easiest, and most profitable for Sun to constantly reinforce the equation: Sun = Sparc = Solaris = Solid
Aiding the development of other OSes leads to...
Sun = Sparc = Another processor, with lower MHz than Intel.
Now on the Intel side, there are two factors at work I figure. First of all is the fact that through purchases and blunders, they're moving into it with boxes like the LX50. Given that fact, they (a) want to get Solaris on as many machines as possible, and (b) want to keep their toes in the Linux waters. Add to that, the fact that when they tried to kill of Solaris/x86, there was a large backlash.
So on the Intel side, they develop Solaris and Linux both. Developing SunLinux is a safety measure which in the short term will sell a few more systems to die-hard Linux admins, while developing Solaris/x86 will keep Solaris on machines that people couldn't justify the cost of Sparc gear for.
OK, so this is all rambling. What it boils down to is this: Sun, like most companies, says "We don't sell computers--we sell SOLUTIONS!" Well on the enterprise side of things, companies don't buy computers--they buy solutions. Buying a PC from the guy down the street, installing Linux, configuring IPTables, locking it down, etc. etc. is not as appealing for most companies as buying an LX50/Solaris/FW1 box and having a single vendor for complete support.
Or to summarise the summary, (nearly) NONE of those copies of Solaris/x86 that Sun sells for $20 will go onto serious production machines--the sort of machines that Sun sells and supports. They'll all end up on hobbiest machines, family web servers, and tiny corporate LANs. This isn't enterprise computing, and it's not going to affect Sun's bottom line.
Re:Why do the fathers of UNIX dislike Linux so muc
on
Dennis Ritchie Interviewed
·
· Score: 3, Insightful
Sigh. I don't know if this is cute or sad.
' "Pieces that are good and pieces that are not." What project is this NOT true of?'
Um...Unix? Unix V.6? Have you read the source for it? It's brilliant! It's occasionally scary and convoluted, but it's GOOD CODE! Linux is an unholy mishmash of some good code, and some deeply sucky code that barely works at all. Sendmail is good code. BIND is good code.
Linux taken seriously on the non-X86 platform is being left to the manufacturers. The various ports of Linux to Sparc processors, for instance, most definitely do NOT hold up! SunLinux will hold up, because it's being developed by Sun for Sun.
Why is it that whenever anyone points out some of the valid and legitimate problems with (a)Linux, or (worse!!!) (b) Open Source development as a philosophy they're categorised as either Microsoft apologists or obsolete? (depending on their respect in the *ix world)
Here's a trick. Go back and reread the article. Think about WHY Ken would say what he did. Think about WHY Dennis Ritchie is fairly unimpressed with Linux. Think about WHY Bill Joy is frightened by the future of computing. Quit dismissing them as 'old farts from before I was born' and you might learn something. God forbid, you might even become a better developer/admin/geek.
Heh. You had me going for the first paragraph, since I was already going to post about the stupidity of listing peak power-handling capabilities as a measure of anything at all except...peak power handling capability.
OK, let me cut down to what I see as the single point of contention.
"Basically you are saying that if an emplpoyer can show that there is risk to others then certain things become justifiable."
No, that's not what I'm saying. That was an example. What I'm saying is that if there is a CAUSAL LINK between behaviour and increased risk, then it's a fair question.
It's not a question of who the risk is to, or even how great it is. It's a question of a definite link between the behaviour (bad credit, drinking, etc.) and the chance of it affecting the person's performance. That link had better be near 100% before you start asking questions, in my mind.
Keep in mind that I'm not proposing legislation here. However if I get questions like this from a prospective employer, I'll walk away before taking the job.
"I think a company should be able to require almost anything before hiring someone."
I think a company *IS* able to require almost anything before hiring someone. I don't think they SHOULD be able to do so.
Mandatory drug tests. Credit checks. How about sexual preference screening? After all, gay men have a higer suicide rate than straight men, not to mention the health care costs of AIDS patients. Can't hire higher risk groups!!!
NONE of these are valid for most jobs. I could get totally stoned on the weekend and it wouldn't necessarily affect my performance. I could be broke and in debt, and it wouldn't necessarily affect my performance.
Companies should NOT be allowed to screen for anything other than a very minimal set of direct, causal situations for a given job. Commercial pilots for instance, SHOULD be screened for alcohol and drugs. I would have no problem, as a commercial pilot, filling in reports on recent consumption before every single flight. At 6000 ft air pressure, tiny amounts of mental dullness magnify.
But unless you're part of the financial services sector, you shouldn't be subject to a credit check. It's stupid, it's invasive, it accomplishes nothing, and it's immoral.
Bah. OK, IRQs were a really bad example. I'm actually running 98SE on one box (because 2k blows it up), and 2k on everything else. My family is all running 2k or XP.
How about if I said I bypass people asking me to fix their devices which don't work or don't install or whateverthefuck doesn't work in Win2k. There ARE still problems--lots of problems--but I will acknowledge that a stable win2k environment is remarkably stable.
"If I switch, the Microsoft area of my brain will atrophy and I'll won't be able to answer the tech questions and assistance asked of me by my friends and family."
This is EXACTLY why I use Windows as little as possible. After dealing with (relatively) heavy iron Sun equipment all day, I get really TIRED of having the family and in-laws ask me about how to fix their IRQ conflicts. Lately I've started to tell them that unless it's Unix I can't help them, and to an increasing extent, it's true. I don't know how Windows goes together, and I don't really care. I'll use it to play my games, and that's it.
Slashdot Mirror
One point.
If it comes to a war? The US has over 50,000 troops over there! There's no way they're going to bring the boys home without giving them something to shoot at first.
You just answered your own question.
The US has *NOT* done a good job of dealing with bin Laden. He's still alive, al Qaeda is still active, anti-US sentiment is at an all-time high in the middle east. What better way to distract people from the failure of dealing with al Qaeda than a war they're guaranteed to win with someone who the US public knows as a Bad Guy? Hussein has thumbed his nose at the US so many times that he has a callous on it, and yet he hasn't really done anything in GWB's regime to justify a war. That alone is probable cause for a war.
The whole "weapons of mass destruction" argument is such a sick joke that it gives me hives. Tell me again which is the ONLY country in the history of the human race to use nuclear weapons against another country? Tell me which country has the largest stockpile of weapons of mass destruction? Oh that's right, it's the holier-than-thou US.
North Korea is violating UN resolutions left, right, and centre; but the US is still happily supporting them. The US is promising to tear up the 1972(?) international nuclear non-proliferation treaty and also start developing chemical/biological weapon programs, but they don't see anything wrong with THEMSELVES doing it. No, it's only when an arrogant Enemy Dictator(tm) with huge oil reserves (another important point) gives them half an excuse to start a guaranteed-to-win war which will distract the population from how ineffective they've been elsewhere that they MUST go to war for the good of the world.
How's that?
OK, we can't retire the space shuttle today. Nor tomorrow. But its time is drawing near...
Consider first that the shuttle was a massive compromise versus the original proposed designs. If the budget had been infinite, we would have had a better shuttle. If the budgeteers had had more foresight, we might (probably) have had a better shuttle. The shuttle we have now is a big series of compromises that limit its usefulness and safety.
Now consider that the shuttle program has been around since 1981. That's more than half of the time that's passwd since man first walked on the moon! It still seems shiny to some of us (myself included), because it was the only newsmaking bit of space exploration in our youth. However, it's old. It's an old (and limited) design, and we have learned a lot of what to do (or not) on the next go around. It's time to climb the next step of astronautical evolution.
So let's keep them in top shape, fly them as necessary (mostly as ferries to the ISS), while putting as much money as possible into a next-generation space vehicle.
I don't really understand why you would say that Solaris is hard to administer. The documentation is fantastic--head and shoulders above most other manufacturers out there. Solaris is an elegant system V OS, with a handful of BSD tools retained (in addition to the Sv ones) For those who want 'em.
Admintool gone? Thank god! That was unquestionably the WORST thing that Sun ever did. The Management console is better by far, although as you point out, almost unusably slow thanks to Java. Sigh.
Moving to Linux tools would be a HUGE step backwards. XFree86 config easier? Xfree86 is the single worst Unix/Linux related tool of any sort, anywhere, anyhow, that has ever been created in my opinion! I have spent over a hundred hours (!!!!!) of my time trying (often unsuccessfully) to get Xfree86 working on different boxes. It is ugly, it is horrible, it is buggy. It is BAD!!! Sun X? It just works. Turn on the computer and go.
vim and GNU grep? No point. There's nothing added in vim other than eye candy, bugs, and poor documentation. GNU grep has a few nice features, but nothing that's much better than egrep, which comes standard with Solaris, HP-UX, and AIX.
Linux has a few simple problems that you're trying to drive Solaris into embracing here. Linux suffers EXTREME and CONSTANT versionitis. The stupid dependencies and constant upgrades are silly and unnecessary. This is exacerbated by the multiple distros of Linux out there, which aren't even particularly compatible with each other, let alone with de facto Unix standards. To pick on one tool, xinetd is a disastrous mess. Any tool that replaces something that's been in existence for 20+ years, and yet doesn't maintain backwards compatibility is entirely unacceptable.
Then we get to the documentation, or lack thereof. There are no doc standards, no consistent format, no guarantee of version-concurrency. The number of times I've seen in a Linux man page, "this document is not kept up to date" is just sick. The info docs aren't much better, and the how-tos are worse. Informal docs written and maintained (or not) by any random volunteer who can type is not how you write documentation.
The TCO of Linux is currently lower, because the only people who run Linux servers are knowledgable Linux-heads. If a Linux solution was purchased by a company for a professional, controlled, change-managed, production environment, then they'd see their TCO SKYROCKET vs. Solaris, or HP-UX, or AIX, or even IRIX.
On the Linux report card, there's a comment from the teacher: "Doesn't play well with others, suffers from inflated self-worth, disrespectful of playmates." Dumping a good solution for a hackers OS because YOU understand it better is a bad idea. A company spending lots of money needs more than just a robust OS (which Linux, mostly, is): They need a consistent, stable (from a development/update point of view), coherent, compatible solution. Turning all of those solutions into Linux is a bad idea.
It's very simple, and can be summed up in one word.
RUN!!!
Escape while you can! You only THINK you want a woman in your life. Don't do it, nooooooo!!!
Bah! cvs SUCKS!!!
OK, cvs sucks in the particular instance of serving up relatively static files. There should be a requirement that everyone who sets up an inappropriate cvs server (where ftp or http would be a better idea) should have to write "versions are not files" on a blackboard, by hand, in chalk, 1000 times. I keep seeing sites that have final, stable, production releases of program 'x,' and the only way to get it is cvs. Dumbness.
Now for actual version concurrency/control, cvs rocks and rocks hard!
Grrr.
1) He already stated this fact. 40% I/O throughput increase. (actually quite a large variability, but it's a usable number)
2) Read the subject line! He said, "THE AVERAGE POWER USER..." Now I read that as meaning the home Linux geek/developer who likes messing with the guts of their system. Companies use RAID0 all the time, or more often RAID1+0. RAID5 is equally common, implemented in hardware. This is not what he's talking about. This is not the target of his comment.
Sigh. Sorry to rant, but every follow up to this article has neglected this point.
I'm not trying to elevate him on a pedestal here, but I just don't want to see him vilified. He screwed up. Big deal. He flipped out on ONE comment out of several. Big deal. He had the decency to apologise. Not a huge deal there either. None of this is a big deal. He seems like a decent guy. That's all.
:-)
On the other hand, I'd like to see someone squeeze an apology out of John Romero.
Good. On. Mark. Rein.
He admitted that they screwed up. (or fucked up, as the case may be.) He lost it when pivx when public. Then he apologised for losing it, and admitted that pivx was entirely in the right.
This is about as much news as the bug itself. Not much.
Bah. Don't bother with the smelly little troll. Let him crawl back into his biometrically-secured bunker, and wait for the rest of the world to forget about him.
I only have one issue with your post. I'm not sure the US has had all that much goodwill since the late-1950s. Even so, George W. certainly is burning it as fast as he can.
"They aren't, HP was first. You can have x86 blades and PA-RISC blades in the same chassis."
Are you sure about this? The only HP blades I can find are the proliants, which means Compaq, which means intel. I've also been told repeatedly that while both HP and IBM COULD have had non-x96 blades, they don't yet.
I'd love to see it. And your idea of a generic blade frame is something I've thought about too. What a nice idea that would be, but I don't see it happening.
Actually, they announced it a few months ago. Now they've finally given it some substance, although it's hard to follow.
:-) then the computers themselves will know to reallocate more resources to the accounting, as it's more mission critical than gaming. Similarly, it will allow for metered semi-automatic allocation of resources. Imagine being a CIO getting a page and an email one night: "Your webserver is running at 95% capacity, and serving 'x' pages per second. For $20,000, you can double your capacity. Please reply with "yes" or "no" in the subject line, to enable this feature or not as you want." Neat plan, eh? It's something that Sun talks about at any rate. It's feasible, with some work.
:-)
N1 is a low-administration, self-managing environment. It's a farm of servers and infrastructure that can dynamically reallocate themselves as needed for different loads, and do so intelligently. If your accounting server is getting hammered at month end, but so is your after-hours game server (hey, it's possible!
DISCLAIMER: Since most of this is extrapolation from what I heard, I can't imagine that anything here violates my NDA, but in the off chance that it does, it's a lucky guess.
I agree entirely.
:-) At that time, the blade servers they announced led me to think of two things:
Two weeks ago, I saw a private announcement of this technology that they talked about today. (signed the NDA and everything!
1) This heterogenous blade system is INCREDIBLY cool!!! You can have Intel blades, Sparc blades, encryption blades, network caching/acceleration blades, etc. etc. etc. all in one frame. The fact that Sun (late as they are) is the first out of the blocks with this idea is remarkable.
2) I WANT CC-NUMA ON BLADE SERVERS! With all the mysterious hype Sun has given N1 (learning from MS, anyone?), I really hope that this is part of their upcoming plans. One OS instance across "n" computers, where n is a variable. Time will tell, I guess.
"Code is GOOD if it provides users with the features they need."
Hmm. I'm not sure I quite agree with this. I think we're using the word 'code' to refer to two different things.
A PRODUCT (application, tool, whatever) is good if it provides users with the features they need. CODE is good if it's well written. Code is well written if it's clean, logical, consistent, efficient, and extensible. Implied in there is that it's easy to debug as well. (which is actually a result of good code)
The old Unix programs lacked features because they weren't possible, realistic, and/or desired. When graphics hardware didn't exist, there was no need for a GUI-based editor. The fact that those applications are now obsolete doesn't make the code any weaker in the slightest! Sometimes the natural lifespan of a program comes to an end.
Think about this as a tech manual. A first edition of the Unix System Administrator's guide isn't all that useful anymore. It talks about how to build a kernel on OSes that don't exist anymore, and it's such specialised knowledge that it's not particularly transferable to modern Unices. It's obsolete. It's not relevant. The only way to 'fix' it is rewrite it from the ground up.
However, the writing style is as good now as it was then. Nemeth et al raised the bar for Unix manuals, by writing an excellent, well organised, understandable, readable, and entertaining book. None of those qualities are changed by obsolescence.
As for sendmail, it's effectively the same answer. No matter how it's written, sendmail is not the right tool for every mail job! There are very good reasons for writing smaller, lighter, easier mail MTU. There are also reasons for keeping sendmail. None of this is remotely relevant to the coide quality of any of them. If a sledgehammer is so good at its job, why would anyone waste their time inventing a finishing hammer? Or a ball-pein? They're DIFFERENT TOOLS!
"GNU versions of fileutils, for instance, corrected a lot of these limitations, at the cost of uglified source code"
Again, I disagree. the GNU tools certainly did fix a lot of limitations in the original tools (and those limitations in some cases are the result of bad code I'll admit, or at least poor choices for code), but the uglified source is not a necessary cost. There's no reason that they can't be written as elegantly as the original ones, or possibly better. From what I've seen of the current version of BIND, it's still clean well written code. There is nothing inherent about modern, 'complex,' massive computing that forces code to be ugly. Only laziness and poor practices on the part of individual programmers causes that. On a massive project, the only way to prevent that sort of bad code is to rigorously screen the programmers, and rigorously screen the code. (Note: Rigorously here means a formalised mandatory code review--having the code available for people to read and fix isn't sufficient.) Unfortunately, OSS by its very nature lacks in both of these areas. I think that was, in a roundabout way, the point that Ken was trying to make.
"Sun = Sparc = I don't care = $$$$"
This is a neat theory. It's not how business works in computing.
Why buy Sun? Why would ANYONE buy Sun from the equation above? My equation had a 'take to the purchasing manager' result, and that is SOLID. "I don't care" (assuming you mean about what OS is run on it) isn't something that makes business decisions. "I don't care" doesn't sell computers. Flexibility of OSes on Sparc is COMPLETELY irrelevant. In fact, nobody really cares what Unix variant is on a box when they're buying it. They know that when they buy Sparc, they're buying into Solaris, and that combination should be unbreakable. If they weaken that link, THEN their sales will suffer.
And x86 just isn't the same market. It never has been, and it never will be. Nothing that Sun does on the x86 side will substantially affect their Sparc sales. The Intel stuff is neat and perhaps profitable, like the SunRays, but it's not an integral part of their business, and it won't sway anyone towards or away from buying a Sparc.
Many reasons.
1) People are lazy.
2) People feel they owe their employer nothing, and have no loyalty. (perhaps partly justified)
3) The modern work ethic seems to be show up in the morning, go home in the late afternoon, and make sure your work all gets done. Whether you do that at work or at home, and when you show up or go home (within reason) seems to be less and less concrete.
If I'm given a three hour job at 4:00 that MUST be done today, and am not getting overtime for it, then be damed if I make sure my coffee breaks don't go over 15 minutes. Alternatively, if I'm on a punch clock and my workday is over at 5:00 then I'm going to work my ass off until that very moment.
I think ultimately the workplace has become more relaxed, which gives slackers and whiners more room to complain without getting their lazy asses fired. The rest of us will just enjoy having more freedom in how we get our work done. (and work more effectively for it)
"Maybe at some point state and Federal regulators will realize that the Bells are the problem, not the solution."
OK, I don't know enough about the phone situation in the US to comment on the subject, but is this line REALLY necessary? I mean it's flat out incitement and misdirection--personal opinion masquerading as factual content.
You're missing out on one of Sun's biggest points: reliability.
On the Sparc platform size, Sun builds high-end, high-reliability, high-predictability boxes. They want an OS that works (a) very well with the hardware, and (b) with the same reliability features. If they're going to promote porting of other OSes onto their platform, they'll do it on their own terms, and with their own requirements, and that's not very straightforward.
It's smartest, easiest, and most profitable for Sun to constantly reinforce the equation:
Sun = Sparc = Solaris = Solid
Aiding the development of other OSes leads to...
Sun = Sparc = Another processor, with lower MHz than Intel.
Now on the Intel side, there are two factors at work I figure. First of all is the fact that through purchases and blunders, they're moving into it with boxes like the LX50. Given that fact, they (a) want to get Solaris on as many machines as possible, and (b) want to keep their toes in the Linux waters. Add to that, the fact that when they tried to kill of Solaris/x86, there was a large backlash.
So on the Intel side, they develop Solaris and Linux both. Developing SunLinux is a safety measure which in the short term will sell a few more systems to die-hard Linux admins, while developing Solaris/x86 will keep Solaris on machines that people couldn't justify the cost of Sparc gear for.
OK, so this is all rambling. What it boils down to is this: Sun, like most companies, says "We don't sell computers--we sell SOLUTIONS!" Well on the enterprise side of things, companies don't buy computers--they buy solutions. Buying a PC from the guy down the street, installing Linux, configuring IPTables, locking it down, etc. etc. is not as appealing for most companies as buying an LX50/Solaris/FW1 box and having a single vendor for complete support.
Or to summarise the summary, (nearly) NONE of those copies of Solaris/x86 that Sun sells for $20 will go onto serious production machines--the sort of machines that Sun sells and supports. They'll all end up on hobbiest machines, family web servers, and tiny corporate LANs. This isn't enterprise computing, and it's not going to affect Sun's bottom line.
Sigh. I don't know if this is cute or sad.
' "Pieces that are good and pieces that are not." What project is this NOT true of?'
Um...Unix? Unix V.6? Have you read the source for it? It's brilliant! It's occasionally scary and convoluted, but it's GOOD CODE! Linux is an unholy mishmash of some good code, and some deeply sucky code that barely works at all. Sendmail is good code. BIND is good code.
Linux taken seriously on the non-X86 platform is being left to the manufacturers. The various ports of Linux to Sparc processors, for instance, most definitely do NOT hold up! SunLinux will hold up, because it's being developed by Sun for Sun.
Why is it that whenever anyone points out some of the valid and legitimate problems with (a)Linux, or (worse!!!) (b) Open Source development as a philosophy they're categorised as either Microsoft apologists or obsolete? (depending on their respect in the *ix world)
Here's a trick. Go back and reread the article. Think about WHY Ken would say what he did. Think about WHY Dennis Ritchie is fairly unimpressed with Linux. Think about WHY Bill Joy is frightened by the future of computing. Quit dismissing them as 'old farts from before I was born' and you might learn something. God forbid, you might even become a better developer/admin/geek.
Heh. You had me going for the first paragraph, since I was already going to post about the stupidity of listing peak power-handling capabilities as a measure of anything at all except...peak power handling capability.
And then...
Heh. Well done.
OK, let me cut down to what I see as the single point of contention.
"Basically you are saying that if an emplpoyer can show that there is risk to others then certain things become justifiable."
No, that's not what I'm saying. That was an example. What I'm saying is that if there is a CAUSAL LINK between behaviour and increased risk, then it's a fair question.
It's not a question of who the risk is to, or even how great it is. It's a question of a definite link between the behaviour (bad credit, drinking, etc.) and the chance of it affecting the person's performance. That link had better be near 100% before you start asking questions, in my mind.
Keep in mind that I'm not proposing legislation here. However if I get questions like this from a prospective employer, I'll walk away before taking the job.
"I think a company should be able to require almost anything before hiring someone."
I think a company *IS* able to require almost anything before hiring someone. I don't think they SHOULD be able to do so.
Mandatory drug tests. Credit checks. How about sexual preference screening? After all, gay men have a higer suicide rate than straight men, not to mention the health care costs of AIDS patients. Can't hire higher risk groups!!!
NONE of these are valid for most jobs. I could get totally stoned on the weekend and it wouldn't necessarily affect my performance. I could be broke and in debt, and it wouldn't necessarily affect my performance.
Companies should NOT be allowed to screen for anything other than a very minimal set of direct, causal situations for a given job. Commercial pilots for instance, SHOULD be screened for alcohol and drugs. I would have no problem, as a commercial pilot, filling in reports on recent consumption before every single flight. At 6000 ft air pressure, tiny amounts of mental dullness magnify.
But unless you're part of the financial services sector, you shouldn't be subject to a credit check. It's stupid, it's invasive, it accomplishes nothing, and it's immoral.
Bah. OK, IRQs were a really bad example. I'm actually running 98SE on one box (because 2k blows it up), and 2k on everything else. My family is all running 2k or XP.
How about if I said I bypass people asking me to fix their devices which don't work or don't install or whateverthefuck doesn't work in Win2k. There ARE still problems--lots of problems--but I will acknowledge that a stable win2k environment is remarkably stable.
Including my firewall and my laptop, I currently have six. Once I take my linux box home, that'll be seven.
OK, so I'm a bit hardcore geek. Having three Sun boxes is a bit uncommon I guess.
"If I switch, the Microsoft area of my brain will atrophy and I'll won't be able to answer the tech questions and assistance asked of me by my friends and family."
This is EXACTLY why I use Windows as little as possible. After dealing with (relatively) heavy iron Sun equipment all day, I get really TIRED of having the family and in-laws ask me about how to fix their IRQ conflicts. Lately I've started to tell them that unless it's Unix I can't help them, and to an increasing extent, it's true. I don't know how Windows goes together, and I don't really care. I'll use it to play my games, and that's it.