Odd. So applying the above to any given amount of RAM would imply it was too much RAM. The logical reduction of this is that the ideal amount of RAM is no RAM at all...
With Quartz 2D Extreme (marketing!) putting the entire rendering of the display onto the graphics card as an OpenGL surface, and lots of the display-rendering code itself being stored there as well, you can never have too much RAM - especially with the composition manager etc. all eating up gobs of it...
It would be nice to see some test results for Apple's GCC versions 3 and 4.
Well, I did have a bunch of results for you, but the CRAPPY LAMENESS FILTER won't let me post them. Apparently I have to use less 'junk' characters (of course the CRAPPY PROGRAMMER didn't define what a 'junk' character is in the error message, so that's NO USE WHAT-SO-EVER.)
So, I guess I'll summarise. gcc version 4 is slightly worse than 3.3, and slightly better when the tree-vectorize option is passed and altivec code is generated.
- Most projectors are 1024x768. This means your HD signal of 1920x1080i or 1920x720p is being heavily rescaled horizontally! Few projectors have a true HD native resolution.
I have a Mitsubishi X390U linked up at home - it's your typical 1024x768 resolution. I've got the comcast HD box linked up to a TiVO (SD only) and (HD feed) directly into the amp's component inputs. The result is that I can switch between HD & SD at the flick of a button. It all gets projected onto an 8' screen.
The difference between HD footage and the *same* SD footage is that of night and day. Heavily rescaled or no - at the end of a movie the tiny credit text is easily readable on HD (as opposed to a blurred just-about-legible mess on the SD feed). The movie itself just blows me away. Every time:-)
Same source, same amp, same projector. World of difference.
So, (and I'm not disagreeing over the scaling), you do get an excellent upgrade from SD when using a 1024x768 projector.
I you're displaying sequential frames (that are *supposed* to be offset spatially) one on top of the other, you'll get aliasing as the images "bounce" up and down. On a moving picture this would just look like blurring, but if you could see them a frame at a time, you'd see something like:
| | | | | (frames) --__--__ (pixels)
(I can't make it line up, but you get the idea...)
I think you'd be better off just dropping odd (or even) fields and upscaling...
See my post above re: the HD DVR. I'm intending to take a feed from the Y,Cb,Cr (or R,G,B) signals that are the end-result *after* all the decoding by the encrypted cable/dish network/whatever. Digitise, compress, store, and play, all in hardware:-)
The Analog Devices input chip I'm planning on using has up to 3 (non-simultaneous) HD inputs of Y,Cb,Cr as well as various SD options. The encoder chip has colourspace conversion of RGB->Y,Cb,Cr and has an RGB overlay input - you'd think they'd designed it for a DVR [grin]
I'm in the initial stages of building one - I've got the video decoder/encoder chips on order from Analog Devices, and I'm going to use a 'soft' CPU in an FPGA for the main cpu. I can use the rest of the FPGA fabric for an ide controller and (initially JPEG, then mpeg2 or H.264) hardware encoder/decoder. My calculations predict ~14 hours of HD on a 400GB drive using JPEG, and ~100 hours when I upgrade it to H.264.
With luck it ought to come to ~6 chips (inc. ethernet) + some discreet components and an HD. Cost ought to be ~$200 when I've finished - don't hold your breath though, it'll take me a while to code up a verilog H.264 design. I've done most of the other verilog components before though.
It's actually (according to the BBC and eso.org) 5x the size of Jupiter, or about half the size of our sun. Calling it a mere planet may be a bit harsh - Jupiter itself is a net producer of energy (radiated = 2x incident, roughly), and it's speculated that this is due to gravity forces. This gas-giant 'planet' is presumably more active gravitationally - perhaps 'proto-sun' or 'failed sun' might be a better description (except that discovering a planet is a far greater acheivement than a tiny star...
It means that binary compatibility will be preserved within the 10.x series from now on, ie: all existing API calls will be preserved as-is.
It doesn't stop them from adding new API calls in 10.{5+} or altering the internals to make them work better/faster. So, no, not deemed perfect, just deemed good enough to be a base to work from.
The article does go into this in some depth (even with pictures, IIRC)...
[sigh] I wasn't really confused - I was trying to point out that there was a gaping chasm between "it makes them worse than The Most Evil Corporation On The Planet", and "they're entitled to do what they did". The reasoning doesn't follow. I was well aware it was Wal*Mart he was calling evil.
I guess you could say that if Wal*Mart don't withdraw books because of the bottom line, then their sole allegiance is to their bottom line. I guess you could say that Apple have applied a more principled outlook - that they will not support (even at their own expense) vendors hostile to them. Which is best will depend on your worldview.
Of course, like everyone else here I've not read the book, so I don't know what caused such offence - it could have been the decision to "publish and be damned" itself, or some content within the books.
As for over-rated. Have you seen Apples stock prices since he came back ? Under-rated more like!
[sigh] I'm a shareholder too. I think the principle is important - that the gutter press don't take over the rest of the publishing world - our worldviews obviously differ.
And I very much doubt anything was done improperly. I would guess SJ called a board meeting and it was voted on fairly and squarely.
As for not carrying THAT book alone - that's a decision for Apple. They took it and they'll bear the consequences - see my point about actions and consequences earlier.
Agreed Steve Jobs ain't Apple. I'd be willing to bet he had a lot more shares than you or I do though - or pretty much anyone actually. I have no idea how many shares he owns, but he's probably the majority shareholder...
I guess I'm a bit confused by the rest of your post - "Apple has every right to pull those books", but you label them worst than The Most Evil Corporation On The Planet because of it. If they're entitled, let 'em. As I said, actions have consequences, and Apple will have their own consequences from their own actions.
And Apple (last I checked) is not a government - if they want don't want to support X, it's entirely up to them. Also, this is not bullying (which is a first-action thing - ie: the agressor is the instigator), this is retaliation (ie: defence) and I for one have no problem with it.
Nope - I'm one of the people rather disgusted by the publisher's actions - see above, but I'd say the same thing if it were Bill rather than Steve.
Despite my tagline, I don't think Bill eats babies. I just think Windows is a pile of crap. Linux (and for me now, OSX) is a far better solution for me - mainly because I rarely have to do spreadsheets or other "business" apps. I'm usually coding, and unix (in whatever flavour) works great for that. Oh, and OSX is the best damn unix workstation I've ever used - and I've used a lot.
Personally I'd be damn annoyed if people started publicising my illnesses, my past and my private life as well. For the second time no less! There's this myth that if you're a public figure you're not entitled to a private life. Bollocks.
Speech is (and IMHO ought to be!) free, and the publishers are well within their rights to go against a man's wishes about his biography. Steve is also well within his rights to tell the publishers that they'll not sell a damn thing in his bookstores from now on.
My sympathies are with the man whose life they're laying bare (irrespective of how they cast it) rather than the money-grabbing publishing house. "Quite surprised" is a laugh as well - they sent the proofs to Apple for approval and were asked to withhold publishing. WTF did they expect ?
One of the things that seems to have been lost along the route to our western democracy is that actions have consequences. I'm made up that the act of publishing this book will cause them financial pain - perhaps it'll be as annoying to them as it obviously is to Steve that they've gone ahead and published. Perhaps it'll make them think twice about doing the same thing again...
Before anyone gets on their high horse about the 'public's right to know', again, Bollocks. The public has a right to know if a public figure abuses his/her position - completely agree with that. On the other hand, this rather distasteful desire to simply nose into other peoples lives is a sad fact of the human condition today.
It was 600 MB/sec, not necessarily one experiment producing 600 MB/sec. Divide and conquer onto multiple RAIDs if you're using linux boxes. Problem solved.
Or use a high-end SGI box or Sun box with multiple SCSI channels bonded into one filesystem with xlv (or the sun equivalent). Each channel goes to a different RAID. Problem solved.
Or... (there are loads of ways to do it, just by dividing the bandwidth up into manageable chunks). Problem solved.
More to the point, the time it would take to get the data onto and off the tapes is left out of your argument. The bandwidth of a truck full of tapes is an old argument, but they're just so damn slow at both endpoints, they're not that useful after all:-(
When the data arrives through a network pipe, it's on disk ready to be crunched through whatever program you're running...
8 or 9 years ago, I used to work in the post-production industry in Soho, London. There's a network called 'Sohonet' where lots of the major post-houses had ATM links to each other (hey, ATM was blazingly fast for the time:-) instead of sending runners with bags full of tapes. It was worth the expense of digging the road and installing the network for them, even then with the slower network.
If a commercial website can't support itself via its audience, that website should die. If the users of the website are sufficiently motivated to pay for content, they will, and it will survive. Here's a hint: if you need to be paid, then be up-front and honest about it (eg: LWN). If your worth preserving, you'll be fine.
There is no such thing as an implied or "social" contract - by their very nature, contracts are not implications! The whole terminology is a marketing exercise designed to appeal to the "guilt" that just because someone is giving you something, you ought to pay for it.
Sheesh! Social contracts! What next ? Breathing contracts ?
You oughtn't need to use a text editor when developing - the Xcode editor is very nice, and imho pretty damn sophisticated... It has code completion, syntax highlighting, click-to-go-to-definition-of-method or declaration-of-function, keyboard shortcuts for code fragment templates, etc. etc. etc. (and I mean etc. - there's a whole load of stuff in there)
Yes, I know you said 'apart from xcode', and I agree (though in my case I'm happy about it) that knowledge of unix is beneficial - I just think your comment over-emphasises the need to know unix... it's not necessary, it just helps the advanced/intermediate programmers.
Whereas I have sympathy for the innocent bystander (as the poster appears to be), and whereas I agree that uncompromising behaviour can be frustrating, the SPAM black hole servers are somewhere between a rock and a hard place...
They can't just block small sections of netblocks (because a spam-happy ISP will just allocate new IP's to their paying spammer customer) - the only way they can police the offence is to ban the block.
They can't just add people back in when they've been blocked either - there has to have been some resolution of the problem, and that has to come from the ISP, at least IMHO. A customer running a website will say anything (especially if they're a scum-of-the-earth-spammer-type customer) to get back online. AN ISP who lies knows their next block will be more permanent...
OTOH, Being unavailable out of hours is... frustrating. In the end, that will reduce the value of the service, and perhaps MAPS will be overtaken by someone who perhaps charges a fee, but is in some what accredited and responsible for their actions.
The real problem though isn't MAPS and their attitude, it's the spammers. Get rid of the spammers and you get rid of the need for MAPS. These lowlife internet-scum are where any ire ought to be directed, again IMHO.
A Sony NDA I once signed said that in the event of disclosure of anything under NDA, Sony would seek damages, and that financial reparation may not be sufficient penalty. The point being that the penalty *ought* to have teeth, and atm, the spam penalties do not. If you want less spam on the 'net, you're going to have to accept more regulation of the 'net. Another double-edged sword...
And a huge percentage of the embedded market is done with those 'pea shooters' . My point is that in the embedded market, speed/RAM size, whatever isn't anywhere near as important. Cost is.
If you wanted a consumer-grade device, there may be an argument, but embedded is (very!) rarely such a resource hog. There is also always a mini-itx type for $100->$200 depending on what you want. Ok, add $40 for 256 MB RAM, and it runs a little slower at (up to) 1GHz but you do get firewire, USB, 100-BASET (up to 4 of them IIRC), TV out, MPEG hardware decoding, LVDS i/o, general purpose i/o (important in embedded), FIR, CIR, audio jacks, and access to more disks...
Or you could pay 2-5 times the price and get a mini (with a DVD-ROM bundled).
... it's a bloody expensive embedded system - most of these are $1-200, not $500. The popular ones are $100 (Rabbit, Arm, AVR, PIC)
Don't get me wrong, I think the Mac mini is just fabulous, but you'd have to have a damn good reason to pay the premium over more traditional embedded systems...
Foreign keys are great when they're necessary. For most web databases (and certainly all the ones I've written for), they'd be useful but are by no means essential - they're there to ensure referential integrity which can be done by error-checking in the caller. At least, that's how I do it.
Stored procedures are (IMHO) a bloody awful idea. Why put the business logic into the database ? All it means is that you have to rewrite the whole thing when you switch DB's. Put middleware where it belongs - in the middle, and use ANSI SQL as much as possible - where you can't abstract it in the middleware. At least that's what I do.
IIRC, stored procedures were introduced to remove the round-trip time between the client and the server, but since both client and server are on the same machine (or gigabit ethernet segment) for all my machines, this isn't a problem, for me. YMMV.
I'm getting a lot of flak on this thread for defending MySQL as a non-ACID database - they typical/. response (if it's not purist-correct, it's a pile of...) is coming out pretty strong. Don't get me wrong, I think ACID will be a great step forward for MySQL - my point was it's a bit off to call a DB not "real" when it's the overwhelmingly most-used DB on the net. It's like calling Windows not a "real" os. Of course it is! Even if it doesn't do what I might like it to do.
The other assumption is that because something is not ACID, it's going to be automatically corrupt - and that's just not the case. I've been running that site at ~100-200 query/sec average for over 5 years (the server's been upgraded - the old one was ~100/sec:-), and every page completely depends on the DB. If the DB was corrupting things it would be very very apparent very quickly. I've written a lot of code to make sure it doesn't get corrupt (and yes, I accept it would be easier if the DB was ACID - you still have to cope with the errors, though ACID or not!)
Slashdot Mirror
Correction: You can always have too much ram
Odd. So applying the above to any given amount of RAM would imply it was too much RAM. The logical reduction of this is that the ideal amount of RAM is no RAM at all...
Simon.
With Quartz 2D Extreme (marketing!) putting the entire rendering of the display onto the graphics card as an OpenGL surface, and lots of the display-rendering code itself being stored there as well, you can never have too much RAM - especially with the composition manager etc. all eating up gobs of it...
Simon
It would be nice to see some test results for Apple's GCC versions 3 and 4.
Well, I did have a bunch of results for you, but the CRAPPY LAMENESS FILTER won't let me post them. Apparently I have to use less 'junk' characters (of course the CRAPPY PROGRAMMER didn't define what a 'junk' character is in the error message, so that's NO USE WHAT-SO-EVER.)
So, I guess I'll summarise. gcc version 4 is slightly worse than 3.3, and slightly better when the tree-vectorize option is passed and altivec code is generated.
Simon
- Most projectors are 1024x768. This means your HD signal of 1920x1080i or 1920x720p is being heavily rescaled horizontally! Few projectors have a true HD native resolution.
:-)
I have a Mitsubishi X390U linked up at home - it's your typical 1024x768 resolution. I've got the comcast HD box linked up to a TiVO (SD only) and (HD feed) directly into the amp's component inputs. The result is that I can switch between HD & SD at the flick of a button. It all gets projected onto an 8' screen.
The difference between HD footage and the *same* SD footage is that of night and day. Heavily rescaled or no - at the end of a movie the tiny credit text is easily readable on HD (as opposed to a blurred just-about-legible mess on the SD feed). The movie itself just blows me away. Every time
Same source, same amp, same projector. World of difference.
So, (and I'm not disagreeing over the scaling), you do get an excellent upgrade from SD when using a 1024x768 projector.
Simon
I you're displaying sequential frames (that are *supposed* to be offset spatially) one on top of the other, you'll get aliasing as the images "bounce" up and down. On a moving picture this would just look like blurring, but if you could see them a frame at a time, you'd see something like:
| | | | | (frames)
--__--__ (pixels)
(I can't make it line up, but you get the idea...)
I think you'd be better off just dropping odd (or even) fields and upscaling...
Simon.
See my post above re: the HD DVR. I'm intending to take a feed from the Y,Cb,Cr (or R,G,B) signals that are the end-result *after* all the decoding by the encrypted cable/dish network/whatever. Digitise, compress, store, and play, all in hardware :-)
The Analog Devices input chip I'm planning on using has up to 3 (non-simultaneous) HD inputs of Y,Cb,Cr as well as various SD options. The encoder chip has colourspace conversion of RGB->Y,Cb,Cr and has an RGB overlay input - you'd think they'd designed it for a DVR [grin]
Simon.
I'm in the initial stages of building one - I've got the video decoder/encoder chips on order from Analog Devices, and I'm going to use a 'soft' CPU in an FPGA for the main cpu. I can use the rest of the FPGA fabric for an ide controller and (initially JPEG, then mpeg2 or H.264) hardware encoder/decoder. My calculations predict ~14 hours of HD on a 400GB drive using JPEG, and ~100 hours when I upgrade it to H.264.
With luck it ought to come to ~6 chips (inc. ethernet) + some discreet components and an HD. Cost ought to be ~$200 when I've finished - don't hold your breath though, it'll take me a while to code up a verilog H.264 design. I've done most of the other verilog components before though.
Simon.
It's actually (according to the BBC and eso.org) 5x the size of Jupiter, or about half the size of our sun. Calling it a mere planet may be a bit harsh - Jupiter itself is a net producer of energy (radiated = 2x incident, roughly), and it's speculated that this is due to gravity forces. This gas-giant 'planet' is presumably more active gravitationally - perhaps 'proto-sun' or 'failed sun' might be a better description (except that discovering a planet is a far greater acheivement than a tiny star...
Simon.
It means that binary compatibility will be preserved within the 10.x series from now on, ie: all existing API calls will be preserved as-is.
It doesn't stop them from adding new API calls in 10.{5+} or altering the internals to make them work better/faster. So, no, not deemed perfect, just deemed good enough to be a base to work from.
The article does go into this in some depth (even with pictures, IIRC)...
Simon
[sigh] I wasn't really confused - I was trying to point out that there was a gaping chasm between "it makes them worse than The Most Evil Corporation On The Planet", and "they're entitled to do what they did". The reasoning doesn't follow. I was well aware it was Wal*Mart he was calling evil.
I guess you could say that if Wal*Mart don't withdraw books because of the bottom line, then their sole allegiance is to their bottom line. I guess you could say that Apple have applied a more principled outlook - that they will not support (even at their own expense) vendors hostile to them. Which is best will depend on your worldview.
Of course, like everyone else here I've not read the book, so I don't know what caused such offence - it could have been the decision to "publish and be damned" itself, or some content within the books.
As for over-rated. Have you seen Apples stock prices since he came back ? Under-rated more like!
Simon.
can I use that in my classroom?
:-)
[grin] feel free
Simon
[sigh] I'm a shareholder too. I think the principle is important - that the gutter press don't take over the rest of the publishing world - our worldviews obviously differ.
And I very much doubt anything was done improperly. I would guess SJ called a board meeting and it was voted on fairly and squarely.
As for not carrying THAT book alone - that's a decision for Apple. They took it and they'll bear the consequences - see my point about actions and consequences earlier.
Simon.
I concede the point :-)
Simon.
I find it disturbing that you do not consider Jobs to be abusive of his position. ;-)
:-)
Had to chuckle
Simon
Agreed Steve Jobs ain't Apple. I'd be willing to bet he had a lot more shares than you or I do though - or pretty much anyone actually. I have no idea how many shares he owns, but he's probably the majority shareholder...
I guess I'm a bit confused by the rest of your post - "Apple has every right to pull those books", but you label them worst than The Most Evil Corporation On The Planet because of it. If they're entitled, let 'em. As I said, actions have consequences, and Apple will have their own consequences from their own actions.
And Apple (last I checked) is not a government - if they want don't want to support X, it's entirely up to them. Also, this is not bullying (which is a first-action thing - ie: the agressor is the instigator), this is retaliation (ie: defence) and I for one have no problem with it.
Simon
Nope - I'm one of the people rather disgusted by the publisher's actions - see above, but I'd say the same thing if it were Bill rather than Steve.
Despite my tagline, I don't think Bill eats babies. I just think Windows is a pile of crap. Linux (and for me now, OSX) is a far better solution for me - mainly because I rarely have to do spreadsheets or other "business" apps. I'm usually coding, and unix (in whatever flavour) works great for that. Oh, and OSX is the best damn unix workstation I've ever used - and I've used a lot.
Simon.
Personally I'd be damn annoyed if people started publicising my illnesses, my past and my private life as well. For the second time no less! There's this myth that if you're a public figure you're not entitled to a private life. Bollocks.
Speech is (and IMHO ought to be!) free, and the publishers are well within their rights to go against a man's wishes about his biography. Steve is also well within his rights to tell the publishers that they'll not sell a damn thing in his bookstores from now on.
My sympathies are with the man whose life they're laying bare (irrespective of how they cast it) rather than the money-grabbing publishing house. "Quite surprised" is a laugh as well - they sent the proofs to Apple for approval and were asked to withhold publishing. WTF did they expect ?
One of the things that seems to have been lost along the route to our western democracy is that actions have consequences. I'm made up that the act of publishing this book will cause them financial pain - perhaps it'll be as annoying to them as it obviously is to Steve that they've gone ahead and published. Perhaps it'll make them think twice about doing the same thing again...
Before anyone gets on their high horse about the 'public's right to know', again, Bollocks. The public has a right to know if a public figure abuses his/her position - completely agree with that. On the other hand, this rather distasteful desire to simply nose into other peoples lives is a sad fact of the human condition today.
Simon.
It was 600 MB/sec, not necessarily one experiment producing 600 MB/sec. Divide and conquer onto multiple RAIDs if you're using linux boxes. Problem solved.
... (there are loads of ways to do it, just by dividing the bandwidth up into manageable chunks). Problem solved.
Or use a high-end SGI box or Sun box with multiple SCSI channels bonded into one filesystem with xlv (or the sun equivalent). Each channel goes to a different RAID. Problem solved.
Or
Simon
More to the point, the time it would take to get the data onto and off the tapes is left out of your argument. The bandwidth of a truck full of tapes is an old argument, but they're just so damn slow at both endpoints, they're not that useful after all
When the data arrives through a network pipe, it's on disk ready to be crunched through whatever program you're running...
8 or 9 years ago, I used to work in the post-production industry in Soho, London. There's a network called 'Sohonet' where lots of the major post-houses had ATM links to each other (hey, ATM was blazingly fast for the time
Simon
If a commercial website can't support itself via its audience, that website should die. If the users of the website are sufficiently motivated to pay for content, they will, and it will survive. Here's a hint: if you need to be paid, then be up-front and honest about it (eg: LWN). If your worth preserving, you'll be fine.
There is no such thing as an implied or "social" contract - by their very nature, contracts are not implications! The whole terminology is a marketing exercise designed to appeal to the "guilt" that just because someone is giving you something, you ought to pay for it.
Sheesh! Social contracts! What next ? Breathing contracts ?
Simon
Yes, I know you said 'apart from xcode', and I agree (though in my case I'm happy about it) that knowledge of unix is beneficial - I just think your comment over-emphasises the need to know unix... it's not necessary, it just helps the advanced/intermediate programmers.
No flames, just a difference of opinion.
Simon
Whereas I have sympathy for the innocent bystander (as the poster appears to be), and whereas I agree that uncompromising behaviour can be frustrating, the SPAM black hole servers are somewhere between a rock and a hard place...
They can't just block small sections of netblocks (because a spam-happy ISP will just allocate new IP's to their paying spammer customer) - the only way they can police the offence is to ban the block.
They can't just add people back in when they've been blocked either - there has to have been some resolution of the problem, and that has to come from the ISP, at least IMHO. A customer running a website will say anything (especially if they're a scum-of-the-earth-spammer-type customer) to get back online. AN ISP who lies knows their next block will be more permanent...
OTOH, Being unavailable out of hours is
The real problem though isn't MAPS and their attitude, it's the spammers. Get rid of the spammers and you get rid of the need for MAPS. These lowlife internet-scum are where any ire ought to be directed, again IMHO.
A Sony NDA I once signed said that in the event of disclosure of anything under NDA, Sony would seek damages, and that financial reparation may not be sufficient penalty. The point being that the penalty *ought* to have teeth, and atm, the spam penalties do not. If you want less spam on the 'net, you're going to have to accept more regulation of the 'net. Another double-edged sword...
Simon
And a huge percentage of the embedded market is done with those 'pea shooters' . My point is that in the embedded market, speed/RAM size, whatever isn't anywhere near as important. Cost is.
If you wanted a consumer-grade device, there may be an argument, but embedded is (very!) rarely such a resource hog. There is also always a mini-itx type for $100->$200 depending on what you want. Ok, add $40 for 256 MB RAM, and it runs a little slower at (up to) 1GHz but you do get firewire, USB, 100-BASET (up to 4 of them IIRC), TV out, MPEG hardware decoding, LVDS i/o, general purpose i/o (important in embedded), FIR, CIR, audio jacks, and access to more disks...
Or you could pay 2-5 times the price and get a mini (with a DVD-ROM bundled).
Simon
... it's a bloody expensive embedded system - most of these are $1-200, not $500. The popular ones are $100 (Rabbit, Arm, AVR, PIC)
Don't get me wrong, I think the Mac mini is just fabulous, but you'd have to have a damn good reason to pay the premium over more traditional embedded systems...
Simon
Foreign keys are great when they're necessary. For most web databases (and certainly all the ones I've written for), they'd be useful but are by no means essential - they're there to ensure referential integrity which can be done by error-checking in the caller. At least, that's how I do it.
/. response (if it's not purist-correct, it's a pile of...) is coming out pretty strong. Don't get me wrong, I think ACID will be a great step forward for MySQL - my point was it's a bit off to call a DB not "real" when it's the overwhelmingly most-used DB on the net. It's like calling Windows not a "real" os. Of course it is! Even if it doesn't do what I might like it to do.
:-), and every page completely depends on the DB. If the DB was corrupting things it would be very very apparent very quickly. I've written a lot of code to make sure it doesn't get corrupt (and yes, I accept it would be easier if the DB was ACID - you still have to cope with the errors, though ACID or not!)
Stored procedures are (IMHO) a bloody awful idea. Why put the business logic into the database ? All it means is that you have to rewrite the whole thing when you switch DB's. Put middleware where it belongs - in the middle, and use ANSI SQL as much as possible - where you can't abstract it in the middleware. At least that's what I do.
IIRC, stored procedures were introduced to remove the round-trip time between the client and the server, but since both client and server are on the same machine (or gigabit ethernet segment) for all my machines, this isn't a problem, for me. YMMV.
I'm getting a lot of flak on this thread for defending MySQL as a non-ACID database - they typical
The other assumption is that because something is not ACID, it's going to be automatically corrupt - and that's just not the case. I've been running that site at ~100-200 query/sec average for over 5 years (the server's been upgraded - the old one was ~100/sec
Simon.
(Phew!)