Sure, they can still use the public computer in the den, they just don't get the _privelege_ of a private computer.
Unless they work, save the money, buy it themselves, and pay for their own phone line. That's what I did when I was _13_. Back then PC-XT 8MHz was common, and 286 12MHz was state-of-the-art, 2400bps was $100, 9600 was $700.
Of course, we didn't have a "family business", weekly allowances, or weird rules about "rights" and "privileges". If any of us could pay for it, that was evidence of enough responsibility to have it.
Worldwide, privatized industries perform better than their state-owned predecessors.
That's actually not true for basic infrastructure industries. Water, wastewater, fire control, and electricity are markets that the government tends to handle far better than private corporations, usually because the corporations try to squeeze as much money out of the system as possible by reducing service and raising prices.
Check out recent stories about the effort to privatize electricity PUCs in the Northeast, or even older stories about what happens when a large city's wastewater operations get privatized.
In fact, most monopoly conditions and competition perversions are, in fact, due to government.
Again, check your history about the rise of monopolies. In most cases, government has to intervene to break them up, resulting in a better deal for consumers. Standard Oil and AT&T being the classic examples.
I'd love to be command-line only, switching to X just for Mozilla, but I've got a couple other X-ish dependencies on the main workstation. When I'm away on the laptop, however, it's Linux CLI all the way.
Reasons why I like the command line *more* than X:
1) The standard Linux VGA font on a 1024x768x16 pixel (128x48 text) VESA console is crisp and easy to read, especially late at night, and especially with color applications like mc, emacs, and tin.
2) The keyboard *always works*. I don't have to fudge with.Xmodmap's for my window manager, emacs, xterm, etc. to get Meta working.
3) gpm's select/paste behaves way better than X's. No worries about missing a character or losing the selecting while clicking between windows.
4) In general, the utilities work more reliably and are faster to use. I remember using tin 1.4 to read Usenet posts that my X-based newsreader would crash on. When running through a long firewall chain, I can use screen to get multiple consoles quickly.
frankly, the simple way to think about this seems the most logical. give the job to the most skilled guy who is willing to do it at the lowest price.
That only works if "price" has a valid meaning, e.g. when currencies are all tied together so that the only reason for the variation in price on a given good is *individual* economic performance of the manufacturer and distribution channel selling that good.
Money markets have gamed the system. Until currencies are locked, any discussion of economics across currency boundaries is wild guesswork.
As recently as 1998, Walmart and other retailers wiped out the camera records 24 hours after they were made. I used to know someone who would write bad checks at Walmart, Target, etc. and then claim the checkbook had been "stolen" the next day. After the police report was filed all of the retailers wrote it off.
The number one deployment platform for IBM software such as DB2 and Websphere remains... Sun.
Interesting. It sounds to me like the IGS reps are probably telling you what you want to hear.
In my experience, IBM's software actually runs "best" (most reliably, and given enough $$$ runs fastest) on pSeries/zSeries. I used to work for Software Group (you can groan if you want -- SWG has its own issues) and we had a lot more hassles using DB2 and WebSphere on Solaris than AIX, Windows, or Linux. Mainly due to IBM's requirement to use the Sun JVM on Solaris -- the Sun JVM enterprise extensions (CORBA, JCE, etc) aren't quite as solid in combination with WebSphere as IBM's version of those libraries.
Don't get me wrong, I didn't see much real *hostility* to Sun hardware inside Big Blue. I'm sure if I was supporting an existing Sun deployment I'd talk about all the happy things that IBM can do there, but if I was selling *new* hardware, I'd try to get some pSeries boxes sold before anything else. I can't imagine that IGS would on the whole be driving ambivalent customers *toward* Sun.
The GP is correct about one thing, though: the default IGS web hosting environment for extremely high-volume web sites is done on IBM hardware.
But still a necessary one! I was looking for a simple program that could do ssh with ANSI color, scrollback, and zmodem (handy for transferring through those pesky firewalls). I couldn't find one, so I started to write my own BBS-era client that could work on today's Internet.
I was a Qmodem 5.x user, hence my project is being made to resemble that interface.
With kids, you don't throw out the rules for sake of convenience or with the idea of being "progressive" about child rearing. The consequences are just too dire.
I'm not trying to slam, but what actually are the consequences of unfiltered net access? Realistically:
1) If a kid is somehow communicating with a real sexual predator, wouldn't their behavior in the outside world show that some kind of problem needs to be addressed?
2) Has anyone shown a solid correlation between childhood/adolescent pornography viewing and adult dysfunction?
I mean, how is Internet access so much different from say unfiltered library access? When I was a lot younger I could easily find dozens of naked pictures in the art and sexology books, and I also found plenty of books undermining my parent's religious, political, and ethical views. I don't feel like I was mentally or morally hurt by that process, so I don't quite understand the need to shield others from it.
In all seriousness, as a parent what are your concerns?
Your finance guy is having the same problem that everyone has with database performance when the row count gets high. Commercial database products do not perform efficiently on many millions of records, no matter what they claim.
That's sort of true, but not entirely. I've worked on DB2 databases that have a few million rows each in several token tables and billions of rows in the star / join tables. With the right indexes in place and of course a schema designed for high volume performance is certainly "reasonable", in which "select * from BIG_TABLE where... " can return in merely seconds.
The problem is that at this size you have to resort to vendor-specific features to keep things chugging along. For instance, you pretty much *have* to insert rows using the custom bulk loader that can temporarily turn off constraint checks. You also need to map the physical layout correctly to the medium, so that database pages are aligned on a good boundary for the SCSI-like IO subsystem.
If you're still curious, there's a great book on DB2 optimization called "DB2 High Performance Design and Tuning". I'm sure Oracle has a similar volume out somewhere.
I was a zealot at one time, but after a few years of real development I came off the high horse and now see both Linux and Windows as different tools for different jobs. But I am clearly in the minority: Microsoft believes that the Windows stack is the right tool for *every* job, and that just isn't so.
Through my jobs over the years I've had dozens of arguments with decision makers who just refuse to even consider the advantages the open-source stack would give them (security, reliability, vendor freedom) and ultimately back their arguments with "I like Windows and that's all there is to it." Even on the servers, they want Windows and will budget to spend big bucks on license fees, even when $0 solutions exist that can be implemented in a similar time frame. My former employer lost some *real money* as lost development time when the worms tore through the network. After a long while you just get sick of saying "I told you so" and it becomes "you got what you deserved".
So I'm personally glad that new zealots are still out there to push the boundaries of polite speech. Some of the decision-makers I've dealt with were hard-headed and didn't argue fair anyway, so I don't feel "we" are diminished by having our own stubborn advocates. And in the meantime, Linux *does* get better, it supports more and more hardware right out of the box, and the X11 desktops get more well-rounded.
The only things missing on my own short list are: professional Linux-based games, clones of critical business applications, and a critical mass of *good* Linux developers. The last bit is of course the hardest and (based on my real experience) *every operating system* needs a lot more *good* developers than are out there now.
For those who REALLY like the command line, here's another terminal emulator that works straight off the raw Linux console: Qodem. Check the README and screenshots page to understand a little why I felt compelled to clone a DOS-based BBS-era program.:)
Currently on the blurry alpha/pre-alpha stage, but another six months and it'll be pretty cool.
Interesting contrast between GP and you. GP is basically asserting that a 40-hour work week is the maximum, and you are responding (correctly BTW) that job commitment still has to be met one way or another.
I have to ask: are you saying that it should be fine for a company to redefine the job *after* the hiring so that more than 40 hours is the new job requirement? That's kind of what it looks like you're saying.
Thanks, you jogged my memory on what I don't like about client-side Java. I'll be adding some text to that section in my own Java problems essay.
Summary: similar experience, but on the server end. Java's a reasonable tool but not the end-all, and anyone who has invested significant time exploring its "cross-platform" claims will see that they are just that: unsupported hype.
Re:Developers, Developers, Developers
on
Mono Beta 2 Released
·
· Score: 0, Offtopic
Clustering (for HA, not for compute tasks) is much more advanced on any decent Unix like AIX or Solaris than with Linux.
True that. I'm used to seeing it in RS/6000 land where two machines are both wired into the same SSA array and one can take over instantly.
The SCSI subsystem in Linux is a joke, and it seems to be even worse with 2.6.
Mmm, why do you say that? I haven't used Linux SCSI much in the high end environment, but I'm on SCSI now and have no problems. My point with the SCSI comment is that for the same amout of cash you spend on a custom Sun array you can get equivalent SCSI hardware that will work with Linux. Of course I'm thinking disk arrays mainly, are you talking the *rest* of SCSI (scanners, tape libraries, etc)?
Does speed matter or doesn't it?;-)
Depends on the application. The one I was most recently paid to work on speed mattered a LOT, but it was all single-CPU bound. Java-based, internal architecture problems galore: ran fastest on Windows 2000. Ick. We couldn't the damn thing to scale very well and no amount of tweaking seemed to help. (More gory details here.)
For my other projects, speed mattered but not so much as availability of source code. For instance, AIX sort is about 3x faster than GNU sort, even when GNU sort was compiled with the native compiler (xlC). But I took the GNU sort.c and re-factored it for my uses (seriously crippling it in the process for general-purpose use) and ultimately got it about 2x faster than AIX sort, BUT with a couple other features specific to our data set I got it really close to the speed of a raw copy (i.e. infinitely fast at sorting).
So your mileage varies on the speed issue. Most times I think the processor more than makes up for the compiler.
Besides that, Solaris used to be a much better system to port software to or from other Unix systems than any other. AIX is maybe not the worst one, but many system features work pretty much different than on other Unix branded systems.
I agree with you there. I've also seen bits of HP-UX, OSF1, and Dynix, and Solaris was definitely closer to BSD and GNU-ish systems in those days (1993-ish). Today I would have to give any one of the free BSDs the crown as "best Unix to port to" due to their use of libc over glibc. (glibc's quiet use of calloc() instead of malloc() has caused me some serious headaches.) From inside C land, AIX 4.x and earlier is a strange beast indeed. But with AIX 5L you get really close to a decent GNU-ish development system.
Still though, my original point is that Linux already has good coverage of the 95% case. Getting that last 5% is of course quite difficult, but I think the existing open-source movement will get there on their own much faster than doing something with the Solaris code.
I'm not excessively familiar with Solaris from an admin standpoint, but I have done quite a bit of porting C/C++ stuff to it and a lot of admin'ing AIX. To be blunt, I don't care much for Solaris and should I ever be in the position to authorize a purchase I'd almost certainly look at Linux first and AIX second. Here's why:
1) Linux is pretty darn good. It would take some *unusually* serious needs before you *have* to look outside the Linux camp to find a workable solution. Linux has XFS, JFS, and ReiserFS, really good support for reliable and fast high-end SCSI, SMP, Beowulf'ing, and a huge community to provide free-as-in-beer help.
2) On a per-processor basis, Linux-on-Intel/PowerPC is faster than Solaris-on-anything hands down. (This will probably change after the next generation of Sparc chips comes out.)
3) Solaris tends to be a pain to port code to. Much like AIX, it's got the AT&T-derived libraries and proprietary crud that doesn't function with as much polish as the GNU stuff. So you end up installing a huge set of GNU tools and libraries on Solaris and... geez by this time you've almost got GNU/Linux again on Sun hardware. AIX 5L has at least started to reverse the trend -- you can get most of the GNU tools pre-installed. (Yes, the native compiler on Solaris and AIX produces much faster code than gcc. Most of my apps don't need the speed, they need the portability. I can optimize at the higher layers and get the speed I need.)
I see plenty of places where *today* Solaris has a great role, but I don't see much in the future. And Sun hardware is nice, but certainly not extraordinarily better than IBM hardware.
This just seems like "too little too late". (Of course, this leads right into the critical question: is there *anything* Sun can do that would be worth paying for?)
I wonder what the world would be like if my grandparents were still around and healthy and vibrant as say.. 40 year olds?
I wonder too. My grandparents (and aunts/uncles/mother) were/are racist homophobic "Christian" bigots who consistently vote to increase the misery in the world. Having seen them bicker with each other for the last couple decades I'm somewhat relieved that my future children won't have to fight the exact same (impossibly stupid) battles I did.
Let the Greatest Generation go to their rest. We've got enough work to do cleaning up their mess.
Interesting. I'm from Texas, and have never heard of this rule. Explains some of the behavior I saw in North Carolina though.
Seriously though, what's the point of it? It looks to me only to encourage bad driving habits. Emergency vehicles have flashing lights and loud sirens and we know very well to get the hell out of the way ASAP. (I've even run through red lights to evacuate the left turn lane so they can pass.) I've seen a couple times a car speeding with hazards on and lights flashing and it's obvious that an emergency situation is taking place, so I move.
But forcing people to get out of the left lane, and restricting trucks out of the left lane (popular along I-40), just seems to encourage bad drivers to zip into the left lane and cause a pile-up on the inside of the highway where it's that much more difficult for the rescue vehicles to get in and control the scene.
Out of curiosity, is there some other rule from the east/west coast states mandating that drivers should get up on an interstate highway ramp and slow to a complete stop? Never saw that in Houston -- though Houstonians hate using their damn blinkers to indicate a lane change, at least they accelerate up to the traffic speed before merging in.
Correct, you cannot circumvent a lawsuit over a patent unless you reach a license agreement before the suit starts. If you go to court, it's basically a toss-up whether you'll win or lose. If you lose you pay damages. But if you lose AND the judge thinks you did so deliberately then you pay triple damages.
You may read all the patents that you *think* cover similar techniques, but there's another one out there that's been filed but not yet issued so you don't see it, then you get sued, and simply because you were trying to be vigilant against stepping on anyone's patent you pay three times. If you are small-to-medium sized you've just gone out of business.
Many software companies on the "innovative" side have a policy in place forbidding employees to read any patents at all to avoid this possibility. As in if you read patents for "fun" you'll get fired, because the risk is too great: if a single employee of your firm has read a patent at the time you lose the suit, you pay triple.
I saw two separate patents related to the same technique. I'm not sure where the novelty is claimed to be. It's not the simple fact that a parameter is on the SRC tag since that has a lot of prior art; I *think* it's described as a method to collect the logfile data even when a caching proxy is in the way. The original filing date is somewhere around 1999.
*I* never filed for patents inside IBM. I'd probably have about five by now (~ $3000) if I had, but I like to be ethical.
I found that browsers were cacheing them, so it wouldn't always register if it was viewed in a webmail acount.
PATENT ALERT
I am about to describe a patented technique. Seriously. If you ever think you're going to implement a web bug, do not read this or IBM will be able to sue you for treble damages.
Since a) I no longer work for IBM, and b) the method is on file in the patent, I am not violating my IP contract with IBM by describing this method.
. . .
PATENT ALERT
. . .
Method:
The way to defeat browser caching is to make the IMG SRC point to a CGI that returns a REDIRECT (302) that points to the single-pixel image. So you might have IMG SRC="server/path/to/cgi?key1=val1&key2=val2". The browser will have to tick the CGI because it has "dynamic" parameters. However, the CGI has to return a REDIRECT because an intelligent proxy server in the middle might be trying to cache the output too. You don't care if the single-pixel image itself is cached, you just want to capture the CGI hit with all the parameters.
My view is that you are hugely exaggerating what 'business' programming is; I feel you are saying that its anything but technical uses?
Well, mainly I mean that 'business' programming is focused on the problem of automating the movement of dollars, inventory, or items around database tables. Computers are basically just "input ==> function ==> output", but business programming is really restricted beyond that. It's "fun" to see where SOAP and EJB can take you (smoothly spanning multi-node systems on the front OR the backend), but after a few years you notice that you haven't:
Implemented a cool algorithm
Used a floating point number
Drawn on a real screen, or played a sound through some speakers...etc
It's just a big data pump to replace the paperwork of clerical staff. That's what I see all these jobs advertising ("web site developer $55K!" "database application developer $65K!" etc.).
That's the crux of our argument really: Java as it's really used today versus what it's claimed to be used for. *I* think Java is primarily being used to first move backend data around and provide human interface to business data, and second for scientific crunching. The games are neat and probably going to eventually penetrate via cell phones, but I think it'll be a long while before Java is featured in a $100 million console game.
Java *is* a good general-purpose language for the companies paying for internal software development (you're dead on with the mention of an organization seeking a re-usable code library BTW -- that's a good point). It's a good sweet spot of what we know works well in a programming language: OOP, compile-time exception checking, and garbage collection. (You can get similar features working in Perl or C++, but it's harder.)
I like the problems Java is trying to tackle, but I think it needs more time to reach critical mass outside the banks and e-tailers. I can certainly understand your perspective -- seeing the industry fragment Smalltalk, C++, Unix, and every other good idea to come along. We finally have an OOP language that all the big iron vendors are willing to standardize on and yet can still be used at home.
My experience was going through school during the early 90's games boom (iD headquarters was 30 miles away) and going through Pascal, C, Visual Basic, Kornshell, Perl, C++, and Java. By the time I got to C++ garbage collection and pthreads were available, and every compiler knew how to do templates. So going to Java I see more what's missing (not yet duplicated) than what is there.
Slashdot Mirror
Sure, they can still use the public computer in the den, they just don't get the _privelege_ of a private computer.
Unless they work, save the money, buy it themselves, and pay for their own phone line. That's what I did when I was _13_. Back then PC-XT 8MHz was common, and 286 12MHz was state-of-the-art, 2400bps was $100, 9600 was $700.
Of course, we didn't have a "family business", weekly allowances, or weird rules about "rights" and "privileges". If any of us could pay for it, that was evidence of enough responsibility to have it.
Worldwide, privatized industries perform better than their state-owned predecessors.
That's actually not true for basic infrastructure industries. Water, wastewater, fire control, and electricity are markets that the government tends to handle far better than private corporations, usually because the corporations try to squeeze as much money out of the system as possible by reducing service and raising prices.
Check out recent stories about the effort to privatize electricity PUCs in the Northeast, or even older stories about what happens when a large city's wastewater operations get privatized.
In fact, most monopoly conditions and competition perversions are, in fact, due to government.
Again, check your history about the rise of monopolies. In most cases, government has to intervene to break them up, resulting in a better deal for consumers. Standard Oil and AT&T being the classic examples.
I'd love to be command-line only, switching to X just for Mozilla, but I've got a couple other X-ish dependencies on the main workstation. When I'm away on the laptop, however, it's Linux CLI all the way.
.Xmodmap's for my window manager, emacs, xterm, etc. to get Meta working.
Reasons why I like the command line *more* than X:
1) The standard Linux VGA font on a 1024x768x16 pixel (128x48 text) VESA console is crisp and easy to read, especially late at night, and especially with color applications like mc, emacs, and tin.
2) The keyboard *always works*. I don't have to fudge with
3) gpm's select/paste behaves way better than X's. No worries about missing a character or losing the selecting while clicking between windows.
4) In general, the utilities work more reliably and are faster to use. I remember using tin 1.4 to read Usenet posts that my X-based newsreader would crash on. When running through a long firewall chain, I can use screen to get multiple consoles quickly.
frankly, the simple way to think about this seems the most logical. give the job to the most skilled guy who is willing to do it at the lowest price.
That only works if "price" has a valid meaning, e.g. when currencies are all tied together so that the only reason for the variation in price on a given good is *individual* economic performance of the manufacturer and distribution channel selling that good.
Money markets have gamed the system. Until currencies are locked, any discussion of economics across currency boundaries is wild guesswork.
Pretty much anywhere south of Mason-Dixon line.
As recently as 1998, Walmart and other retailers wiped out the camera records 24 hours after they were made. I used to know someone who would write bad checks at Walmart, Target, etc. and then claim the checkbook had been "stolen" the next day. After the police report was filed all of the retailers wrote it off.
The number one deployment platform for IBM software such as DB2 and Websphere remains... Sun.
Interesting. It sounds to me like the IGS reps are probably telling you what you want to hear.
In my experience, IBM's software actually runs "best" (most reliably, and given enough $$$ runs fastest) on pSeries/zSeries. I used to work for Software Group (you can groan if you want -- SWG has its own issues) and we had a lot more hassles using DB2 and WebSphere on Solaris than AIX, Windows, or Linux. Mainly due to IBM's requirement to use the Sun JVM on Solaris -- the Sun JVM enterprise extensions (CORBA, JCE, etc) aren't quite as solid in combination with WebSphere as IBM's version of those libraries.
Don't get me wrong, I didn't see much real *hostility* to Sun hardware inside Big Blue. I'm sure if I was supporting an existing Sun deployment I'd talk about all the happy things that IBM can do there, but if I was selling *new* hardware, I'd try to get some pSeries boxes sold before anything else. I can't imagine that IGS would on the whole be driving ambivalent customers *toward* Sun.
The GP is correct about one thing, though: the default IGS web hosting environment for extremely high-volume web sites is done on IBM hardware.
It all seems such a distant past now..
But still a necessary one! I was looking for a simple program that could do ssh with ANSI color, scrollback, and zmodem (handy for transferring through those pesky firewalls). I couldn't find one, so I started to write my own BBS-era client that could work on today's Internet.
I was a Qmodem 5.x user, hence my project is being made to resemble that interface.
With kids, you don't throw out the rules for sake of convenience or with the idea of being "progressive" about child rearing. The consequences are just too dire.
I'm not trying to slam, but what actually are the consequences of unfiltered net access? Realistically:
1) If a kid is somehow communicating with a real sexual predator, wouldn't their behavior in the outside world show that some kind of problem needs to be addressed?
2) Has anyone shown a solid correlation between childhood/adolescent pornography viewing and adult dysfunction?
I mean, how is Internet access so much different from say unfiltered library access? When I was a lot younger I could easily find dozens of naked pictures in the art and sexology books, and I also found plenty of books undermining my parent's religious, political, and ethical views. I don't feel like I was mentally or morally hurt by that process, so I don't quite understand the need to shield others from it.
In all seriousness, as a parent what are your concerns?
Your finance guy is having the same problem that everyone has with database performance when the row count gets high. Commercial database products do not perform efficiently on many millions of records, no matter what they claim.
... " can return in merely seconds.
That's sort of true, but not entirely. I've worked on DB2 databases that have a few million rows each in several token tables and billions of rows in the star / join tables. With the right indexes in place and of course a schema designed for high volume performance is certainly "reasonable", in which "select * from BIG_TABLE where
The problem is that at this size you have to resort to vendor-specific features to keep things chugging along. For instance, you pretty much *have* to insert rows using the custom bulk loader that can temporarily turn off constraint checks. You also need to map the physical layout correctly to the medium, so that database pages are aligned on a good boundary for the SCSI-like IO subsystem.
If you're still curious, there's a great book on DB2 optimization called "DB2 High Performance Design and Tuning". I'm sure Oracle has a similar volume out somewhere.
I was a zealot at one time, but after a few years of real development I came off the high horse and now see both Linux and Windows as different tools for different jobs. But I am clearly in the minority: Microsoft believes that the Windows stack is the right tool for *every* job, and that just isn't so.
Through my jobs over the years I've had dozens of arguments with decision makers who just refuse to even consider the advantages the open-source stack would give them (security, reliability, vendor freedom) and ultimately back their arguments with "I like Windows and that's all there is to it." Even on the servers, they want Windows and will budget to spend big bucks on license fees, even when $0 solutions exist that can be implemented in a similar time frame. My former employer lost some *real money* as lost development time when the worms tore through the network. After a long while you just get sick of saying "I told you so" and it becomes "you got what you deserved".
So I'm personally glad that new zealots are still out there to push the boundaries of polite speech. Some of the decision-makers I've dealt with were hard-headed and didn't argue fair anyway, so I don't feel "we" are diminished by having our own stubborn advocates. And in the meantime, Linux *does* get better, it supports more and more hardware right out of the box, and the X11 desktops get more well-rounded.
The only things missing on my own short list are: professional Linux-based games, clones of critical business applications, and a critical mass of *good* Linux developers. The last bit is of course the hardest and (based on my real experience) *every operating system* needs a lot more *good* developers than are out there now.
For those who REALLY like the command line, here's another terminal emulator that works straight off the raw Linux console: Qodem. Check the README and screenshots page to understand a little why I felt compelled to clone a DOS-based BBS-era program. :)
Currently on the blurry alpha/pre-alpha stage, but another six months and it'll be pretty cool.
Interesting contrast between GP and you. GP is basically asserting that a 40-hour work week is the maximum, and you are responding (correctly BTW) that job commitment still has to be met one way or another.
I have to ask: are you saying that it should be fine for a company to redefine the job *after* the hiring so that more than 40 hours is the new job requirement? That's kind of what it looks like you're saying.
Thanks, you jogged my memory on what I don't like about client-side Java. I'll be adding some text to that section in my own Java problems essay.
Summary: similar experience, but on the server end. Java's a reasonable tool but not the end-all, and anyone who has invested significant time exploring its "cross-platform" claims will see that they are just that: unsupported hype.
OT: an interesting discussion indeed at your sig: What an eye opening dialogue
Seriously, good job: you dropped the bomb and the replies just proved your point. I hope the situation described ultimately worked out in your favor.
Clustering (for HA, not for compute tasks) is much more advanced on any decent Unix like AIX or Solaris than with Linux.
;-)
...or just whinig that Sun is evil. ;-)
;-)
True that. I'm used to seeing it in RS/6000 land where two machines are both wired into the same SSA array and one can take over instantly.
The SCSI subsystem in Linux is a joke, and it seems to be even worse with 2.6.
Mmm, why do you say that? I haven't used Linux SCSI much in the high end environment, but I'm on SCSI now and have no problems. My point with the SCSI comment is that for the same amout of cash you spend on a custom Sun array you can get equivalent SCSI hardware that will work with Linux. Of course I'm thinking disk arrays mainly, are you talking the *rest* of SCSI (scanners, tape libraries, etc)?
Does speed matter or doesn't it?
Depends on the application. The one I was most recently paid to work on speed mattered a LOT, but it was all single-CPU bound. Java-based, internal architecture problems galore: ran fastest on Windows 2000. Ick. We couldn't the damn thing to scale very well and no amount of tweaking seemed to help. (More gory details here.)
For my other projects, speed mattered but not so much as availability of source code. For instance, AIX sort is about 3x faster than GNU sort, even when GNU sort was compiled with the native compiler (xlC). But I took the GNU sort.c and re-factored it for my uses (seriously crippling it in the process for general-purpose use) and ultimately got it about 2x faster than AIX sort, BUT with a couple other features specific to our data set I got it really close to the speed of a raw copy (i.e. infinitely fast at sorting).
So your mileage varies on the speed issue. Most times I think the processor more than makes up for the compiler.
Besides that, Solaris used to be a much better system to port software to or from other Unix systems than any other. AIX is maybe not the worst one, but many system features work pretty much different than on other Unix branded systems.
I agree with you there. I've also seen bits of HP-UX, OSF1, and Dynix, and Solaris was definitely closer to BSD and GNU-ish systems in those days (1993-ish). Today I would have to give any one of the free BSDs the crown as "best Unix to port to" due to their use of libc over glibc. (glibc's quiet use of calloc() instead of malloc() has caused me some serious headaches.) From inside C land, AIX 4.x and earlier is a strange beast indeed. But with AIX 5L you get really close to a decent GNU-ish development system.
Still though, my original point is that Linux already has good coverage of the 95% case. Getting that last 5% is of course quite difficult, but I think the existing open-source movement will get there on their own much faster than doing something with the Solaris code.
Sun ain't evil, they just look evil on TV.
I'm not excessively familiar with Solaris from an admin standpoint, but I have done quite a bit of porting C/C++ stuff to it and a lot of admin'ing AIX. To be blunt, I don't care much for Solaris and should I ever be in the position to authorize a purchase I'd almost certainly look at Linux first and AIX second. Here's why:
... geez by this time you've almost got GNU/Linux again on Sun hardware. AIX 5L has at least started to reverse the trend -- you can get most of the GNU tools pre-installed. (Yes, the native compiler on Solaris and AIX produces much faster code than gcc. Most of my apps don't need the speed, they need the portability. I can optimize at the higher layers and get the speed I need.)
1) Linux is pretty darn good. It would take some *unusually* serious needs before you *have* to look outside the Linux camp to find a workable solution. Linux has XFS, JFS, and ReiserFS, really good support for reliable and fast high-end SCSI, SMP, Beowulf'ing, and a huge community to provide free-as-in-beer help.
2) On a per-processor basis, Linux-on-Intel/PowerPC is faster than Solaris-on-anything hands down. (This will probably change after the next generation of Sparc chips comes out.)
3) Solaris tends to be a pain to port code to. Much like AIX, it's got the AT&T-derived libraries and proprietary crud that doesn't function with as much polish as the GNU stuff. So you end up installing a huge set of GNU tools and libraries on Solaris and
I see plenty of places where *today* Solaris has a great role, but I don't see much in the future. And Sun hardware is nice, but certainly not extraordinarily better than IBM hardware.
This just seems like "too little too late". (Of course, this leads right into the critical question: is there *anything* Sun can do that would be worth paying for?)
Comments?
I wonder what the world would be like if my grandparents were still around and healthy and vibrant as say.. 40 year olds?
I wonder too. My grandparents (and aunts/uncles/mother) were/are racist homophobic "Christian" bigots who consistently vote to increase the misery in the world. Having seen them bicker with each other for the last couple decades I'm somewhat relieved that my future children won't have to fight the exact same (impossibly stupid) battles I did.
Let the Greatest Generation go to their rest. We've got enough work to do cleaning up their mess.
Interesting. I'm from Texas, and have never heard of this rule. Explains some of the behavior I saw in North Carolina though.
Seriously though, what's the point of it? It looks to me only to encourage bad driving habits. Emergency vehicles have flashing lights and loud sirens and we know very well to get the hell out of the way ASAP. (I've even run through red lights to evacuate the left turn lane so they can pass.) I've seen a couple times a car speeding with hazards on and lights flashing and it's obvious that an emergency situation is taking place, so I move.
But forcing people to get out of the left lane, and restricting trucks out of the left lane (popular along I-40), just seems to encourage bad drivers to zip into the left lane and cause a pile-up on the inside of the highway where it's that much more difficult for the rescue vehicles to get in and control the scene.
Out of curiosity, is there some other rule from the east/west coast states mandating that drivers should get up on an interstate highway ramp and slow to a complete stop? Never saw that in Houston -- though Houstonians hate using their damn blinkers to indicate a lane change, at least they accelerate up to the traffic speed before merging in.
As an expanding oppressive regime, no.
As a check on the power of another expanding oppressive regime, absolutely.
Or he uses a compass instead of a ruler.
Correct, you cannot circumvent a lawsuit over a patent unless you reach a license agreement before the suit starts. If you go to court, it's basically a toss-up whether you'll win or lose. If you lose you pay damages. But if you lose AND the judge thinks you did so deliberately then you pay triple damages.
You may read all the patents that you *think* cover similar techniques, but there's another one out there that's been filed but not yet issued so you don't see it, then you get sued, and simply because you were trying to be vigilant against stepping on anyone's patent you pay three times. If you are small-to-medium sized you've just gone out of business.
Many software companies on the "innovative" side have a policy in place forbidding employees to read any patents at all to avoid this possibility. As in if you read patents for "fun" you'll get fired, because the risk is too great: if a single employee of your firm has read a patent at the time you lose the suit, you pay triple.
I saw two separate patents related to the same technique. I'm not sure where the novelty is claimed to be. It's not the simple fact that a parameter is on the SRC tag since that has a lot of prior art; I *think* it's described as a method to collect the logfile data even when a caching proxy is in the way. The original filing date is somewhere around 1999.
*I* never filed for patents inside IBM. I'd probably have about five by now (~ $3000) if I had, but I like to be ethical.
I found that browsers were cacheing them, so it wouldn't always register if it was viewed in a webmail acount.
PATENT ALERT
I am about to describe a patented technique. Seriously. If you ever think you're going to implement a web bug, do not read this or IBM will be able to sue you for treble damages.
Since a) I no longer work for IBM, and b) the method is on file in the patent, I am not violating my IP contract with IBM by describing this method.
.
.
.
PATENT ALERT
.
.
.
Method:
The way to defeat browser caching is to make the IMG SRC point to a CGI that returns a REDIRECT (302) that points to the single-pixel image. So you might have IMG SRC="server/path/to/cgi?key1=val1&key2=val2". The browser will have to tick the CGI because it has "dynamic" parameters. However, the CGI has to return a REDIRECT because an intelligent proxy server in the middle might be trying to cache the output too. You don't care if the single-pixel image itself is cached, you just want to capture the CGI hit with all the parameters.
My view is that you are hugely exaggerating what 'business' programming is; I feel you are saying that its anything but technical uses?
...etc
Well, mainly I mean that 'business' programming is focused on the problem of automating the movement of dollars, inventory, or items around database tables. Computers are basically just "input ==> function ==> output", but business programming is really restricted beyond that. It's "fun" to see where SOAP and EJB can take you (smoothly spanning multi-node systems on the front OR the backend), but after a few years you notice that you haven't:
Implemented a cool algorithm
Used a floating point number
Drawn on a real screen, or played a sound through some speakers
It's just a big data pump to replace the paperwork of clerical staff. That's what I see all these jobs advertising ("web site developer $55K!" "database application developer $65K!" etc.).
That's the crux of our argument really: Java as it's really used today versus what it's claimed to be used for. *I* think Java is primarily being used to first move backend data around and provide human interface to business data, and second for scientific crunching. The games are neat and probably going to eventually penetrate via cell phones, but I think it'll be a long while before Java is featured in a $100 million console game.
Java *is* a good general-purpose language for the companies paying for internal software development (you're dead on with the mention of an organization seeking a re-usable code library BTW -- that's a good point). It's a good sweet spot of what we know works well in a programming language: OOP, compile-time exception checking, and garbage collection. (You can get similar features working in Perl or C++, but it's harder.)
I like the problems Java is trying to tackle, but I think it needs more time to reach critical mass outside the banks and e-tailers. I can certainly understand your perspective -- seeing the industry fragment Smalltalk, C++, Unix, and every other good idea to come along. We finally have an OOP language that all the big iron vendors are willing to standardize on and yet can still be used at home.
My experience was going through school during the early 90's games boom (iD headquarters was 30 miles away) and going through Pascal, C, Visual Basic, Kornshell, Perl, C++, and Java. By the time I got to C++ garbage collection and pthreads were available, and every compiler knew how to do templates. So going to Java I see more what's missing (not yet duplicated) than what is there.