I read of a study done running Fedora Core Five versus a Mac G5 running FC5 Actually, the Mac was running FC4, but all the testing was done using a synthetic benchmarking tool, so I doubt that makes any difference. And the last line of the article states:
Geekbench also isn't able to exploit the eight vector processors on the Cell processor. Any program designed and optimized for the Cell processor should be a lot faster than one designed for a generic processor (like, say, Geekbench). So while the Geekbench results might seem disappointing, keep in mind that Geekbench can't exercise the PlayStation 3 to its full potential. So it's hardly the condemnation of the PS3's performance you make it out to be.
Just remember, where the big bad guys see "little people to be silenced," others see "opportunity." Yeah — the opportunity to silence some little people, to curry favor with the big bad guys...
Great, another site I can't visit when I'm at The Client From Hell. TCFH, among other endearing practices, sets DHCP leases to expire every fifteen minutes. As you might imagine, this makes downloading large patches or updates "eventful"...
sometimes the orphans do no harm except take up space, and it's possible that the faster delete of the parent is preferred.
It's a classic time-space trade-off. I disagree. In a classic time/space trade-off, you don't run the risk of producing incorrect results (e.g., someone runs an aggregate function on the detail records, but some of those are invalid because they don't have a parent). Granted, you could argue that "well, in that case the orphans do harm, so that's not what I'm talking about", but there's no way you can guarantee that someone, somewhere, won't need to do something like that.
I think a better example in your case would be to simply add a "valid" flag (or possibly another value for a status field, if the record already has one). Set the valid flag to 'N' and reap all the invalid records and their children as part of your overnight processing. It's a lot cheaper to check WHERE PARENT.VALID = 'Y' than WHERE PARENT.ID IN (SELECT ID FROM PARENT p2). In fact, if you deleted the parents within the same transaction where you perform the check, you might not even be able to use that logic, depending on your transaction isolation level.
Defeating referential integrity is always a Bad Idea. If you really need the speed increase (say during a mass import), then it's usually better to drop the constraint completely, then add it again afterwards. Import code is pretty much the only place where you'll want to enforce constraints at the application level.
Obviously, terrorists will manipulate this information to drive down the stock of the lower-performing carriers (yes, Cingular, I'm looking at you). Then, once they drop low enough, they'll purchase the company at fire-sale prices and...#5 PROFIT!
Even if the swap device becomes inaccessible, it may still be possible to perform some reasonably orderly recovery and shutdown by permitting access to processes that are still in memory. Whoa, nasty college flashback... Poking around on a PDP-11/34 running V6 Unix, looking for a spool file after my listing didn't show up on the printer... I still don't know what I did, but suddenly the console printer spit out "PANIC: SWAP READ ERROR" and the disks stopped chattering. Some things you never forget...
Anyone else notice the link is pointing to the Democratic Leadership Council? Yeah, sorry, the original story I read was in either CACM or EE Times, but I can't seem to find it (it was in one of those "news bites"-type sections, ISTR). So rather than make an empty assertion, I found some backing evidence (for sufficiently loose values of "evidence") of the policy (or attitude) in general.
Nice catch. I didn't note the DLC link, just "ppionline" in the URL, which didn't mean anything to me.
I know that one of the commandments says that we are not supposed to think we can understand the mind of God Hmm, can't say I'm familiar with that one. No other gods before Me, no graven images, name in vain, keep the Sabbath, father and mother, no adultery/stealing/coveting/murder/lying (is that ten?). Nothing about attempting to interpret or understand God's actions. There's the old cop-out "God works in mysterious ways", often used to excuse some event that presumably a God such as the one advertised in the Bible wouldn't have allowed to happen, but nothing in the Commandments (if I've remembered them correctly, no guarantee there...)
The executive branch has some control over what gets researched, and I'm basically OK with that; I would be if it was done fairly, or at least rationally. Refusing to fund a US$30B fusion reactor because the money isn't available is understandable, refusing to permit a prominent US engineer to participate on an international standards committee because he made a donation to a political party other than the one currently occupying the White House is not. Yet this is just what the current administration is doing.
And does (can?) anyone offer enterprise-level Debian? Canonical offers "enterprise-level" support (whatever that means, apparently guaranteed response times [note that "You don't generally see behavior like that in an enterprise-grade appliance" is, technically, a response]). But has anyone certified its performance on specific hardware? I see that HP offers Debian support "packs" (as well as installation assistance), but doesn't certify it on any of their servers.
Maybe I'm off track here, but I could certainly understand MySQL not wanting to offer an enterprise-level product for a platform that wasn't also enterprise-level.
Installation/Deployment guides *are* required documentation. IT takes the software, and follows the guide. And note that these are guides, not step-by-step instructions. They should say things like "Load the database schema update script (app_schema_updt_1_1_19.sql) into the database". The actual mechanics of doing this (making a backup, bringing an additional transaction log file on-line, starting table-level auditing, whatever) are left to the people who actually know the system (and the department procedures). Part of the development process is circulating a draft of the installation guide to IT for comments. IT is free to ask "Why?" or "How?" or "Where am I supposed to get the disk space for that?". I always made a point of inviting someone from IT to our team meetings whenever we discussed deployment requirements, and that always worked well. IT got a heads-up as to what was coming down the pike, and often the IT person had good insight into alternatives and what resources were really available (e.g., "XYZ division got their own server, so all their information in the production system is static. They wanted it updated hourly, but the update locks those tables for over five minutes, so we're just doing it noon and midnight for now." Need I mention that our app was reporting against XYZ division's data, and this was the first we'd heard of it?).
That's interesting. I never considered Sprint because they don't seem to have good coverage away from the major interstates.
I'm about to switch from Cingular to Alltel, because Cingular's GSM network is pretty lame where I live. I considered Verizon and probably would have gone with them, but their family plans run US$10/month higher than the equivalent plans from other carriers.
It's strange: I've been month-to-month with Cingular for probably about a year now. When I asked about a new contract, they offered to extend my current contract and give me a discount on a new phone. When I pointed out I could move to another carrier, keep my existing number and get a couple of new phones, they offered to throw in a cheap phone if I bought a mid-priced one. Hmmm, two Moto V3m RAZRs for free, or a free Samsung C417 if I buy a Moto V3i RAZR (which I don't want) for $79, hmmm, what to do, what to do...
How did everyone here begin learning / teaching themselves about different aspects of programming, that they initially had no clue about? How did you improve? Programming on your own? Through work? A little of both. What I used to do was to pick something I was interested in (say AJAX-enabled web sites), then try to do a project for work using that tech (the "through work" part). They were side projects I worked on during lunch or after normal work hours (the "on your own" part). Then, when I had something that worked, I'd show it to my manager. Usually they'd dismiss it with a "that's interesting, but we really don't need a web page to indicate who's in the office and who's sick/off-site/at lunch/on vacation". Which was fine, I didn't really want to document or maintain these things, but knowing that I was going to demo it gave me a little added incentive to not gloss over some of the details.
Lately, I find that lots of new tech comes with a tutorial project, so I start with that, then flesh it out into something real. If you want to learn Java/JEE, the Pet Shop app is a good place to start, likewise Sun's Blueprints series. I think they're a step above a random SourceForge project because they're designed to be easy to follow. Most of them aren't inherently useful once they're done, though, so you might find it easier to pick up a real project and play with that, since it's useful to start with, and you can work to improve it instead of just learning a coding technique with a project that ultimately doesn't do anything.
Lots of good info and (often) code snippets at specific tech web sites. Searching for "PHP programming site" brings up some good ones. Another thing to try is "XYZ (example | tutorial)" where XYZ is something you're interested in.
If you're really interested in something (or really think it's something you should learn about), try to find a local user group. Probably not practical for anything other than languages or specific products, though (think.NET or MySQL).
on IBM mainframes 30 years ago, booting OS/VS1 under VM/370 took over five minutes ISTR that booting V7 UNIX on a PDP-11 took several minutes as well (a large part of that being the RL02 spin-up). However, the same system running under SIMH boots in roughly a second...
Too bad we can't we have an emulator that emulates your actual machine. It'd report all your (virtual) devices as ready, so at least the initial boot would go quickly. I know some work's being done with cacheing startup files too, but it seems to break down fairly quickly in practice -- you aren't just reading, you're writing log files and configuration settings and other things that eventually have to be written back to the disk once its up.
It might not be a quick as restoring a hibernation image of a freshly-booted system, but I've never heard how those schemes handle things like providing the proper date and time[0], or a machine that's on a network other than the one it's configured for in the boot image (esp, if the configured network has a static IP, and the machine's on a network that requires DHCP). +++ [0] Especially if you've got some kind of monitoring program that grabs the system time on startup and logs all messages with a time delta. You'd wind up with a log with the first date being the snapshot date, and all the deltas being several million seconds later, when the code read the real clock...
I didn't just read the summary, I took a look at the standard.. <xs:complexType name="CT_MetadataRecord">
<xs:annotation>
<xs:documentation>Metadata Record</xs:documentation>
</xs:annotation>
<xs:attribute name="t" type="xs:unsignedInt" use="required">
<xs:annotation>>
<xs:documentation>Metadata Record Type Index</xs:documentation>>
</xs:annotation>
</xs:attribute>
<xs:attribute name="v" type="xs:unsignedInt" use="required">>
<xs:annotation>
<xs:documentation>Metadata Record Value Index</xs:documentation>
</xs:annotation>
</xs:attribute>>
</xs:complexType>
So, we wind up with tags like <CT_MetadataRecord t=4 v=0>. Guess it would have ballooned the spec out if they would have had attributes with actual names like "typeNdx" and "valueNdx"...
What's also interesting is that MS will be offering a "bridge" (as a separate download) that enables Office software to read and write ODF (the OpenOffice Open Document Format) files.
ECMA just confirmed the MS Open Office XML format as a standard, not Office in general. MS further states that OOXML will be an "open and royalty-free" specification.
What's also interesting is that MS will be offering a "bridge" (as a separate download) that enables Office software to read and write ODF (the OpenOffice Open Document Format) files.
is Google part of the Internet, or is the Internet now a subset of Google? Or more generically, is the internet the network, or the services it provides? To many people, the internet is just the WWW. They may know their e-mail travels over it, but to them they "get on" the internet by firing up their browser.
So tell me again why I'd want to continuously RENT my applications rather then buy software licenses and install/run the stuff on my OWN equipment? Maybe because the application has pretty hefty hardware requirements? I notice that Salesforce.com is raking in the dough, largely because most CRM systems require two or three servers (and I don't mean Linux on a white box, think somethinglikethis. And that's per site, you'll probably have your main servers and a second set at a backup site (or at least one big one that can virtualize any of the others). And then there's the bandwidth, power, cooling, storage...
Here's a bad analogy for you: computers used to be like trains — nobody owned their own, you paid to ride someone else's. Eventually, cars became affordable, so most people bought a car instead of taking the train. The situation now is as if someone built this great mass transit system (the internet), and now most people can just dial a number on their phone and a shuttle shows up at their door. Sure, the people with sports cars and classic cars and people who just enjoy driving will keep their cars, but the rest of the folks will be glad to get rid of the maintenance hassle/expense and turn the garage into a media room.
In support of your rental analogy, I think it's more akin to people who lease cars instead of buying them. You get all the benefits of having a car, but at the end of the lease you trade it in for a new model. That's the benefit of renting your applications: you run the latest code and don't have to worry about upgrading, that's handled for you. Likewise with bug fixes. [0] Presumably there's some kind of support for custom work, I haven't really played much in the software services arena. But for a lot of things, it makes sense. Some ISPs already install a standard software bundle with a browser and e-mail client, why not just ship the browser and offer everything else on-line? They could even offer expanded services (like the basic office apps that Google is building), all served from their system.
I'm not convinced this is in "the near future", but it's going to become a trend. And I think a lot of people will jump on it. +++ [0] Yeah, I can figure how much hassle it will be when your app vendor decides your bug is a low priority, or they decide to eliminate some feature you rely on. I think we'll find that vendors who keep their customers happy stay in business, and the market will demand a certain level of service/accountability.
What, exactly, is the internet's competition? Internet II? Minitel? I notice that no-one's offering me a discount to switch from using the internet to "MegaCorp's NEW ULTRAnet! (now with 30% more fiber!)".
I mean, why does Access requesting data from a network database freeze up the entire machine... Damn, guys.. do that crap in a background thread Good sign of a "mature code base" — it's so nasty, they'd rather put a workaround in the OS to support its bad habits than fix it. I remember under NT, Access would peg the CPU at 100% just sitting there. I assume it was actively polling for events it needed to service, probably due to some deficiency in the Win95 message handler (or more likely, to provide some feature). Anybody else remember how Word for the Mac wasn't available under (System 7.5?) for a long time, because apparently the same code base was used for both Windows and the Mac, but the Mac translator/runtime was so nasty and hacked that it took forever to get it working?
And I suspect most of this is caused by a failure to abstract away enough of the OS, for performance reasons. They wind up tying the software to a particular implementation of a feature, so now subsequent versions of Windows have to provide some kind of hook or wrapper to support it. And when it's a major cash cow like Access or Outlook, they'll do whatever it takes to keep it working.
Also, knowing how to code an algorithm is +- useless. What is usefull is understanding it without a language context, and more importantly, its tradeoffs in a given situation. Not to get all Zen or anything, but if you can't code the algorithm, are there any tradeoffs? If all you've got is a nail, why bother to debate the relative merits of Philips vs. Robertson-head screws? You're not going to be using either one of them...
I think they should learn computer languages in the order that they evolved: assembler first, then FORTRAN, then Pascal (as the ALGOL representative), then C, then Smalltalk. Maybe throw Lisp in there between FORTRAN and Pascal (even though I believe it's older than FORTRAN). Actually, C is kind of a step backwards from Pascal (limited scoping and no built-in I/O), maybe just tack that on to the last couple of weeks of assembler.
And let's not forget that students should really be studying computer science, not programming. They shouldn't learn how to do a binary search, they should learn why it's such a powerful technique. The implementation falls out naturally from the description. Likewise for trees of various flavors. Teach them how to identify the language features that best support the algorithms they need, and let them figure out which language is most appropriate for themselves. After all, any language they learn in college will be out of favor by midpoint in their career anyway (or earlier, for those of us that learned Pascal...), so better to teach them how to learn a new one from the start.
Slashdot Mirror
Great, another site I can't visit when I'm at The Client From Hell. TCFH, among other endearing practices, sets DHCP leases to expire every fifteen minutes. As you might imagine, this makes downloading large patches or updates "eventful"...
It's a classic time-space trade-off. I disagree. In a classic time/space trade-off, you don't run the risk of producing incorrect results (e.g., someone runs an aggregate function on the detail records, but some of those are invalid because they don't have a parent). Granted, you could argue that "well, in that case the orphans do harm, so that's not what I'm talking about", but there's no way you can guarantee that someone, somewhere, won't need to do something like that.
I think a better example in your case would be to simply add a "valid" flag (or possibly another value for a status field, if the record already has one). Set the valid flag to 'N' and reap all the invalid records and their children as part of your overnight processing. It's a lot cheaper to check WHERE PARENT.VALID = 'Y' than WHERE PARENT.ID IN (SELECT ID FROM PARENT p2). In fact, if you deleted the parents within the same transaction where you perform the check, you might not even be able to use that logic, depending on your transaction isolation level.
Defeating referential integrity is always a Bad Idea. If you really need the speed increase (say during a mass import), then it's usually better to drop the constraint completely, then add it again afterwards. Import code is pretty much the only place where you'll want to enforce constraints at the application level.
Obviously, terrorists will manipulate this information to drive down the stock of the lower-performing carriers (yes, Cingular, I'm looking at you). Then, once they drop low enough, they'll purchase the company at fire-sale prices and...#5 PROFIT!
Some things you never forget...
Nice catch. I didn't note the DLC link, just "ppionline" in the URL, which didn't mean anything to me.
Great! Now we can make whiskey!
And does (can?) anyone offer enterprise-level Debian? Canonical offers "enterprise-level" support (whatever that means, apparently guaranteed response times [note that "You don't generally see behavior like that in an enterprise-grade appliance" is, technically, a response]). But has anyone certified its performance on specific hardware? I see that HP offers Debian support "packs" (as well as installation assistance), but doesn't certify it on any of their servers.
Maybe I'm off track here, but I could certainly understand MySQL not wanting to offer an enterprise-level product for a platform that wasn't also enterprise-level.
IT takes the software, and follows the guide. And note that these are guides, not step-by-step instructions. They should say things like "Load the database schema update script (app_schema_updt_1_1_19.sql) into the database". The actual mechanics of doing this (making a backup, bringing an additional transaction log file on-line, starting table-level auditing, whatever) are left to the people who actually know the system (and the department procedures). Part of the development process is circulating a draft of the installation guide to IT for comments. IT is free to ask "Why?" or "How?" or "Where am I supposed to get the disk space for that?". I always made a point of inviting someone from IT to our team meetings whenever we discussed deployment requirements, and that always worked well. IT got a heads-up as to what was coming down the pike, and often the IT person had good insight into alternatives and what resources were really available (e.g., "XYZ division got their own server, so all their information in the production system is static. They wanted it updated hourly, but the update locks those tables for over five minutes, so we're just doing it noon and midnight for now." Need I mention that our app was reporting against XYZ division's data, and this was the first we'd heard of it?).
That's interesting. I never considered Sprint because they don't seem to have good coverage away from the major interstates.
I'm about to switch from Cingular to Alltel, because Cingular's GSM network is pretty lame where I live. I considered Verizon and probably would have gone with them, but their family plans run US$10/month higher than the equivalent plans from other carriers.
It's strange: I've been month-to-month with Cingular for probably about a year now. When I asked about a new contract, they offered to extend my current contract and give me a discount on a new phone. When I pointed out I could move to another carrier, keep my existing number and get a couple of new phones, they offered to throw in a cheap phone if I bought a mid-priced one. Hmmm, two Moto V3m RAZRs for free, or a free Samsung C417 if I buy a Moto V3i RAZR (which I don't want) for $79, hmmm, what to do, what to do...
Lately, I find that lots of new tech comes with a tutorial project, so I start with that, then flesh it out into something real. If you want to learn Java/JEE, the Pet Shop app is a good place to start, likewise Sun's Blueprints series. I think they're a step above a random SourceForge project because they're designed to be easy to follow. Most of them aren't inherently useful once they're done, though, so you might find it easier to pick up a real project and play with that, since it's useful to start with, and you can work to improve it instead of just learning a coding technique with a project that ultimately doesn't do anything.
Lots of good info and (often) code snippets at specific tech web sites. Searching for "PHP programming site" brings up some good ones. Another thing to try is "XYZ (example | tutorial)" where XYZ is something you're interested in.
If you're really interested in something (or really think it's something you should learn about), try to find a local user group. Probably not practical for anything other than languages or specific products, though (think
Too bad we can't we have an emulator that emulates your actual machine. It'd report all your (virtual) devices as ready, so at least the initial boot would go quickly. I know some work's being done with cacheing startup files too, but it seems to break down fairly quickly in practice -- you aren't just reading, you're writing log files and configuration settings and other things that eventually have to be written back to the disk once its up.
It might not be a quick as restoring a hibernation image of a freshly-booted system, but I've never heard how those schemes handle things like providing the proper date and time[0], or a machine that's on a network other than the one it's configured for in the boot image (esp, if the configured network has a static IP, and the machine's on a network that requires DHCP).
+++
[0] Especially if you've got some kind of monitoring program that grabs the system time on startup and logs all messages with a time delta. You'd wind up with a log with the first date being the snapshot date, and all the deltas being several million seconds later, when the code read the real clock...
I didn't just read the summary, I took a look at the standard..
<xs:complexType name="CT_MetadataRecord">
<xs:annotation>
<xs:documentation>Metadata Record</xs:documentation>
</xs:annotation>
<xs:attribute name="t" type="xs:unsignedInt" use="required">
<xs:annotation>>
<xs:documentation>Metadata Record Type Index</xs:documentation>>
</xs:annotation>
</xs:attribute>
<xs:attribute name="v" type="xs:unsignedInt" use="required">>
<xs:annotation>
<xs:documentation>Metadata Record Value Index</xs:documentation>
</xs:annotation>
</xs:attribute>>
</xs:complexType>
So, we wind up with tags like <CT_MetadataRecord t=4 v=0>. Guess it would have ballooned the spec out if they would have had attributes with actual names like "typeNdx" and "valueNdx"...
Do you have a source for that?
I read it first here...ECMA just confirmed the MS Open Office XML format as a standard, not Office in general. MS further states that OOXML will be an "open and royalty-free" specification.
What's also interesting is that MS will be offering a "bridge" (as a separate download) that enables Office software to read and write ODF (the OpenOffice Open Document Format) files.
Here's a bad analogy for you: computers used to be like trains — nobody owned their own, you paid to ride someone else's. Eventually, cars became affordable, so most people bought a car instead of taking the train. The situation now is as if someone built this great mass transit system (the internet), and now most people can just dial a number on their phone and a shuttle shows up at their door. Sure, the people with sports cars and classic cars and people who just enjoy driving will keep their cars, but the rest of the folks will be glad to get rid of the maintenance hassle/expense and turn the garage into a media room.
In support of your rental analogy, I think it's more akin to people who lease cars instead of buying them. You get all the benefits of having a car, but at the end of the lease you trade it in for a new model. That's the benefit of renting your applications: you run the latest code and don't have to worry about upgrading, that's handled for you. Likewise with bug fixes. [0] Presumably there's some kind of support for custom work, I haven't really played much in the software services arena. But for a lot of things, it makes sense. Some ISPs already install a standard software bundle with a browser and e-mail client, why not just ship the browser and offer everything else on-line? They could even offer expanded services (like the basic office apps that Google is building), all served from their system.
I'm not convinced this is in "the near future", but it's going to become a trend. And I think a lot of people will jump on it.
+++
[0] Yeah, I can figure how much hassle it will be when your app vendor decides your bug is a low priority, or they decide to eliminate some feature you rely on. I think we'll find that vendors who keep their customers happy stay in business, and the market will demand a certain level of service/accountability.
What, exactly, is the internet's competition? Internet II? Minitel? I notice that no-one's offering me a discount to switch from using the internet to "MegaCorp's NEW ULTRAnet! (now with 30% more fiber!)".
And I suspect most of this is caused by a failure to abstract away enough of the OS, for performance reasons. They wind up tying the software to a particular implementation of a feature, so now subsequent versions of Windows have to provide some kind of hook or wrapper to support it. And when it's a major cash cow like Access or Outlook, they'll do whatever it takes to keep it working.
Yeah, now we know where C-SPAN's Black Ops budget has been going...
I think they should learn computer languages in the order that they evolved: assembler first, then FORTRAN, then Pascal (as the ALGOL representative), then C, then Smalltalk. Maybe throw Lisp in there between FORTRAN and Pascal (even though I believe it's older than FORTRAN). Actually, C is kind of a step backwards from Pascal (limited scoping and no built-in I/O), maybe just tack that on to the last couple of weeks of assembler.
And let's not forget that students should really be studying computer science, not programming. They shouldn't learn how to do a binary search, they should learn why it's such a powerful technique. The implementation falls out naturally from the description. Likewise for trees of various flavors. Teach them how to identify the language features that best support the algorithms they need, and let them figure out which language is most appropriate for themselves. After all, any language they learn in college will be out of favor by midpoint in their career anyway (or earlier, for those of us that learned Pascal...), so better to teach them how to learn a new one from the start.