the max bandwidth of RDRAM is higher than equivalent (PC133 DDR) SDRAM, though latency is higher.
..but that's the deal killer right there!! I'm not playing video games on these machines -- I can't expect linear, consistent reads of large blocks of memory. I have a bunch of unix daemons grabbing all sorts of random chunks from RAM. How is moving from 60 clock ticks for a cache miss to 200 clock ticks a good thing?? Oh, but if I'm moving more than a few K it'll get to the CPU in 25% larger blocks.
This is exactly like Intel pumping up Mhz for marketing purposes rather than actual engineering. High latency, on servers, sucks ass. RDRAM? No thanks.
"If we admire the Net, should not a burden of proof fall on those who would change the basic assumptions that brought it about in the first place?"
I was going to just write off the article as SoCal tissue, but my eye was caught at the end by this:
from the article:
"It's too big, too important, too political to be treated as something for only a band of talented engineers to preside over." -- Michael Roberts, former chair of IANA
*BOGGLE* Oh indeed; lets have those Harvard MBAs configuring Cisco 8100s, shall we? This is a former chair of IANA! He should know better!
Only, as the consoles become more advanced, the companies seem to be losing more on hardware sales. I forget the exact total, but Sony is losing money on every PS2 unit they sell.
From my vauge memory of the figures, I think the loss-per-unit is about the same (adjusted for inflation). One of the oddities of the market, actually, is Nintendo, who won't sell a console until they can make a profit on it -- this is why their consoles come out last (moore's law in effect).
Bottom line is I hate SCEA and I think a majority of the PS library (1st and 3rd party) is crap.
For me, that's just way too personal of a standpoint to takes towards these companies. Besides, I don't have to want the majority of anything - just not enough time or energy to care.
I think this is just a fundamental difference -- myself and my friends hate the idea of a big company telling us what games we're going to like. Hate it so much, in fact, we import odd stuff from Japan when SCEA, normally taking all comers, doesn't think it'll sell. Recent acquisition: 'Vib Ribbon', the only 1.5 dimensional game in existence. Sad thing is, Sega was the same way..
But compared to Sega and Nintendo, Sony's lineup is pretty shallow (989 Studios comes to mind)
Do you mean their in-house lineup? I agree totally; 989 Studios in particular seem to be finely honed masters at the art of sucking.
We seem to be discussing different things. Let me restate/rebut: Nintendo has strong in-house teams because they couldn't trust exterior developers, whereas Sony has, well, jack shit in-house and hordes of good people outside deploying on their hardware. It's also historic: Nintendo is a games company that builds hardware to sell games; Sony is a hardware company that licenses games to sell hardware. Sort of a MS/Sun duality. And I'm happy with the latter.
Sega and Nintendo hold the majority of the "hardcore" crowd.... those of us who'd sit outside ElBo the whole night, blindly waiting for the next Sonic game just because it has Yuji Naka's name on it.
Bah. Hardcore for what? What makes you think what you find cool is what the rest of us find cool? I'm just going to guess that we're the same age (mid-20's); we probably have played many of the same games; we can both agree that Nintendo is powered by Shigeru Miyamoto, a designer as determined and unstoppable as a mecha come to destroy Tokyo. Even so, when was the last time you waited in a line all night? That kid stuff is over -- our generation of gamers now have days jobs and don't mind picking up a game a few days late. This kind of messaniac fervor is passing.
What's ElBo? Must be an East Coast thing...
And while I'm tempted to purchase a PS2 for MGS2, I can't rationalize dropping $300 on a platform for which I know I'm only going to want 3 or 4 games.
OTOH, I bought the PSX for the explicit use of FF7. Money well spent, I should point out (break it down into $ per hour of fun - it's enlightening). I'll buy a PS2, sight unseen, to play FFX. Although I want to play the new Metroid, I know that the only good games will be the ones done by Miyamoto - there will be, what, 4 over the life of the GameCube? (metroid, 2 zeldas, etc?) Not worth it. If Sega's Smilebit team targets it, yeah, that would rule, but they're already targeting PS2, mainly because it's open to all comers.
Anyway, this little discussion just proves my point -- that if two fanatical video game fans with similar backgrounds can't come to agreement on what's important in a game, a big dumb company with a list of trademarks, sales spreadsheets, axes to grind, and hordes of customers spread across many market segments has almost no chance to be able to define, let alone regulate, Quality.
we can develop titles on regular PC workstations, then have a publisher (like Microsoft themselves) foot the minimal cost of the port if the game looks good.
I'm still not sure how well this will work out. Contrary to slashdot belief, the biggest difference between PC and console development is not instruction sets. The distinction is that a PC is a general purpose machine shared between multiple applications; a console is an embedded device. Writing PC code, you want to minimize time and size requirements for efficiency, but you still have a large shared RAM space and even swap when that grows too small. The PC game must also accept the fact that it must share space with cron, IRC, a large OS, whatever. The console is the inverse of both - it has a small, fixed memory, but it doesn't have to share with anyone. So the impetus is to keep the system pegged at all times (churning max polygons; streaming off disc; etc) but it can't exceed the memory & time limits.
This is much more like real-time programming than more conventional desktop coding, and why consoles are harder to work with than PCs. I do not think this problem will magically go away.
When I think of quality games, the Playstation (1 or 2) rarely comes to mind.
Really? Everything Square has done in the last 5 years isn't quality? You can't find quality on a platform that is the unquestioned leader in every market segment that doesn't have the word 'Mario' in it?
How has Sony's quantity over quality creed hurt them? Everyone jumped ship from Nintendo to PSX due to Nintendo's censorship, incredibly restrictive licensing system, and bad technical decisions. Sony's more open licensing system gave them the crown for the last half of the 90's. Sony's one attempt at 'quality control' was trying to restrict RPGs in a misguided attempt at branding the PSX as a action/sport console -- an attempt that hurt everyone involved, none more than Sony. They learned their lesson after seeing which segment sold the most and had the most repeat customers; and now on the PS2 they push the RPGs as the system mainstay.
Games are still a new field, with development very analogous to start-ups. A big dumb company wielding creative control is, and always has been, a dramatic failure. Sony, a big dumb company if ever there was one, has gained a great deal by keeping their hands off their developers.
Quantity, not quality, becuase you can't predict quality.
Asking to speak to a lawyer is the key thing; as it forces their hand to either book charges or else let you go. If they book you, there are then more stringent requirements on evidence, what they can ask you, rules of intimidation, lawyer present, etc; none of which are imposed when you aren't formally arrested. So 'detainment' is a great boon for cops, as it means every word you say can be used to build a case against you, since it's just normal conversation in the eyes of the law.
Oracle is free for development; they only charge when you deploy. It's all there on technet.oracle.com.
Yes, these FrontBase guys are active on comp.database.*, and they're certainly easier to talk to than Oracle developers. For both of those I am glad. But the topic of the 'Ask Slashdot' is something more powerful than Postgres, and FrontBase really isn't even up to that yet, let alone in the range of Oracle and DB2. Stop spamming; read the article; stay on topic.
You're missing the point. The article was nothing more than a pimp rag for Microsoft, who's trying to bust into that market. So of course they're going to ignore anything that suggests they have more advanced technology than ourselves, or that the console-PC dichotomy is real and important.
Americans, South Koreans, and some Europeans play PC games. Everyone else plays console games. If you look closely, the internet stopped being the killer app for PCs a couple years ago, leaving games as the prime mover for faster boxes.
Landline Net access is inhibited by NTT. Thus, Japan has the most advanced cell network in the world, and you can get all manner of net services over DoCoMo. Speaking of innovation, DoCoMo's profit sharing system has created a whole cottage industry of service innovation. To say that software development in Japan is dead is blatantly false.
There's another reason American software companies are having trouble selling in Japan, and it's not just their bad EUC/JIS/Unicode systems. Most US firms have adopted a 'reave and plunder' approach to Japan... Oracle, for instance, charges three times as much for their products as they do here.
Sorry, Economist, but you're shilling. Goodbye, respect.
Re: Overly broad accusations
on
Eco-Terrorism
·
· Score: 3
You just responded to 2 insightful posts by proclaimed environmentalists, both intimately connected to genetics. Both of them decried certain specific failures by large elements of the enviromental movement, but in no way revoked their 'enviromentalist status'. Ergo, your accusations are overly broad, unsupported, and simply bizarre. Particular behaviors in a set do not condemn the whole set.
"environmentalists are... nihilists" environmentalists are the furthest thing from nihilists. We believe that mankind has a future, and the best way to assure that is to moderate our consumption to stay in line with our ability to produce. We're working to provide for our children. Who are you working for?
"Environmentalism is *not* a science." Of course it isn't. It is a political movement, backed by *everything* science has told us over the last 50 years about sustaining humanity's future.
I won't even get into your absurd attempts at psychoanalysis. Please put down the Ayn Rand (a religion if ever there was one) and grow up.
in 99% of cases, a monopoly can only stay a monopoly if they treat the customers reasonably well.
Your argument is totally unsupported and you are taking it as axiomatic, but let's ignore that for a moment. The troubling thing is that you seem to take this as an argument to the corollary that monopolies should not be curbed b/c your theoretical '99% of them' behave well. The latter does not flow from the former.
these are cases where violence and other crimes are used to maintain power. In these situations, it's a different story, and troops move in. Luckily this hasn't happened on any real scale yet, as far as I know.
Hello!!! This situtation is precisely what was happening at the beginning of the 20th century. Read a history book, not a fictional platform for someone's economic theories.
Who in their right mind would store large binary objects into a database anyway? Wouldn't it be much more efficient to have the filesystem to mind about large files and just save filenames and pathes into the DB?
Umm... Oracle, Informix, IBM, MS, Postgres, maybe MySQL, etc, etc. If you go back and read a little more closely, you'll see that DBs aren't storing LOBs in row, they store an address in row that points to a less structured LOB area. Whether this LOB area is in a tablespace controlled by the DB or right on the filesystem is an implementation detail. I don't think you care, but I'll explain a few technical reasons to chose between them.
Most of these systems offer a LOB type that stores the actual LOB in its own file on the filesystem, so you do have that option. The problems with LOB-on-filesystem is that the LOBs are then exposed to, well, any luser on the system, some of whom mistype 'rm' arguments. Besides not being under the control of the DB's security, LOB-on-fs limits locking to standard semaphores (no MVCC), adds a layer of syscalls, complicates caching, and the DB can't do very intelligent reads like it can on its own tablespaces.
Really the only reason to use LOB-on-fs is if you need to access those same objects without going through the DB, like if you want to edit a 500M LOB in emacs. Interestingly, Oracle runs a SMB server in 8i that can expose everything as an apparent filesystem, so even that reason is a bit flimsy....
So AMD might actually find itself some top-class engineers looking for work on its doorstep soon.
Sort of like when DEC was bought by Compaq? We can give credit to the Athlon in part to some key Alpha developers who joined AMD immediately following DEC's quartering. I think the rest of the Alpha engineers know where they'll be welcome...
As an aside, when DEC was bought by Compaq Intel bought DEC's StrongARM and new england foundry. Most of the employees left. From what I hear a goodly chunk of the StrongARM folks joined Cadence and other design firms. I don't know about the rest.
I think there's historical evidence that the only thing Intel is buying here is patents.
I'm watching AMD on this one -- they have a bunch of the original Alpha designers working for them now, so they have the skillz. Aside from them, there are a host of other competition, as you so clearly stated. Some you missed are Hitachi, which ran the Dreamcast and has a growing share of the embedded market; and ARM, which is still kicking out new cores (and working on 64-bit now).
There's an interesting story behind HP end-of-lifeing PA-RISC and SGI eol'ing IRIX; I posted it last night on the previous Alpha story, but maybe it deserves restating:
HP teamed with Intel b/c the CEO-for-hire they had at the time (Rick Belluzio) tried to get the company to drop PA-RISC and HP-UX and move everything to Intel and WinNT (because that's the future!). I wonder in which class in biz school do they tell you to just drop 2 decades of engineering focus and end-of-life all your products at once. The predictable occured: all their server customers went to Sun, who was busy sucking up every internet customer around. Their 'high-end NT workstations' were massively undercut by, well, every PC clone maker in the world. Moral: value your uniqueness. The board managed to fire him in time to reverse some of the damage, but HP was been burned during the fastest growth period for Unix servers in years (possibly ever). They didn't need to before, but now they really need to honor those contracts with Intel.
As an aside, this same CEO-for-hire did the same manuver at SGI (end-of-life MIPS & IRIX, sell WinTel), with the same consequences. He's now working at Microsoft. It's a facinating biz study -- every commercial Unix vendor who partnered with MS & Intel was badly damaged (DEC was destroyed). Sun, who fought MS tooth and nail, thrived. Perhaps it's naive to follow behind WinTel...
It's very funny. Solely as a business decision, these manuvers were supremely moronic. Michael Porter (Harvard B-school professor) defines "strategy" as "doing what is necessary to become unique in your market." Belluzio (& similar folk) took the exact opposite approach, and paid for it. There's the obvious problem of MBAs failing to understand technical specialization, which is a rant that has no end, so I won't touch it. But yesterday a friend and I were talking on similar lines and I gained a new insight: biz school folk are taught 3 or 4 models, they drill those into them, and make sure they know them; but they are never taught tax -- the details of which can wipeout entire industries -- witness the yacht industry in the late 80's. The country is run primarily by biz types and lawyers who don't understand how their models actually apply in the real world.
So I can see that these types 'know' business real well. At least the small subset that they believe is the entirety of biz, and all hell is lose when the map doesn't match the terrain. If there's a moral here, maybe it's "never hire CEOs from outside".
BTW - UT Austin victim right here! Physics department, 99. What's your damage?
In schema definition and SQL parsing, the blob is a normal column. To the database internals, the blob is not stored with the rest of the table, instead a location reference is stored with the normal data that points to a less organized area where LOBs are stored. For simplicity's sake, envision the tabular data as a linear series of 8k disk blocks.(or 4k, or 2k, or whatever). Algorithms scanning the tables can quickly move down this line of blocks, because of the regularity in layout. Such algorithms could not do the same if huge chunks of LOB were in the way, so they are stored elsewhere on disk. Every RDBMS does this, the only differences are the specifics.
What you're missing is that none of this matters to the user -- this is all internal to the database's storage structure, and the great win with the relational model is the separation of physical storage and logical design.
To get back to your specifics, 1) you shouldn't have to have a seperate table for LOBs, 2) Postgres handles LOBs just fine, and 3) you should be wary of streaming BLOBs out of MySQL -- I'm not sure how that goes over given MySQL's table-level locks. I'd suggest ripping the BLOB out quick and caching it in RAM.
Alpha is the only viable competitive 64-bit microprocessor architecture to IA-64 (maybe the last possible mass-produced alternative to Intel-based architectures of any width).
Umm.... Sun's UltraSPARC? SGI's MIPS? HP's PA-RISC? IBM's Power line? A swath of Japanese chips, notably Hitachi? With the exception of SGI, all of these seem to be doing fine. Even ARM is working on 64-bit. To my eye, the 64-bit conversion more or less where the field was before, with the exception of PA-RISC (which will be phased out after McKinley). The literature has been covering 64-bit issues for years now; It's not like one company is holding all the trade secrets. All the niches will remain where they are; the operating systems will carry over (presumably Windows will be 64-bit clean in 3 years). What makes you think IA64 is the end of the world in chip design?
Intel would substantially consolidate the market and remove the largest visible threat to its dominance by taking over Alpha, whatever it did with it afterwards.
Intel's largest visible threat to its dominance is AMD. I can't understand how you're missing that. They have chewed up more of Intel's market share than any competitor ever. And perhaps you aren't aware of this: a bunch of the original Alpha designers work at AMD now. So I'm not scoffing at the promises of their own 64-bit proc.
I just don't see any need for doom-and-gloom here.
HP teamed with Intel b/c the CEO-for-hire they had at the time tried to get the company to drop PA-RISC and HP-UX and move everything to Intel and WinNT (because that's the future!). I wonder in which class in biz school do they tell you to just drop 2 decades of engineering focus and end-of-life all your products at once. The predictable occured: all their server customers went to Sun, who was busy sucking up every internet customer around. Their 'high-end NT workstations' were massively undercut by, well, every PC clone maker in the world. Moral: value your uniqueness. The board managed to fire him in time to reverse some of the damage, but HP was been burned during the fastest growth period for Unix servers in years (possibly ever). They didn't need to before, but now they really need to honor those contracts with Intel.
As an aside, this same CEO-for-hire did the same manuver at SGI (end-of-life MIPS & IRIX, sell WinTel), with the same consequences. He's now working at Microsoft. It's a facinating biz study -- every commercial Unix vendor who partnered with MS & Intel was badly damaged (DEC was destroyed). Sun, who fought MS tooth and nail, thrived. Perhaps it's naive to follow behind WinTel...
Your link leads to a Beowulf integrator. If you work there, you understand the current limits of parallel processing, and the premium companies will pay for really high end equipment. Besides, the hardware is usually the cheapest thing in an organization. And what the hell are you talking about dumping Alpha when IA64 hits the street? Compaq's Alpha business is an 8 billion per year market. Most of those are big corporate customers who won't change platforms for 5 years minimum. Assuming they'll change over this fall presupposes the existence of your magical pet monkey.
Remaining SQL92 issues: subselects in SELECT, subselects in FROM (also called projections), subselects in WHERE. Each can be thought of as a seperate implementation issue. Views. Updateable views. Foreign keys. Constraints, in particular referential integrity constriants. Required transactions - this is part of every SQL standard, it's a base assumption. SELECT INTO table. Temp tables. Domains. Stored procedures and triggers aren't standardized, but every major DB has them, and they're demanded out in the real world -- add them to the list.
After adding all the above, MySQL will need some serious (read: light-years of) development on their query analyzer. MySQL does well on simple selects but performs notably slow on complex queries. Which is hardly a surprise given their assumptions, design goals, and evangelism.
InnoDB needs more testing, and it needs to continue to grow. I have full faith in Heikki Turri's skill.
I cannot say the same for the other MySQL AB developers. They have spent the last few years stonewalling efforts to turn MySQL into a real DB. They proclaimed that transactions were bad, table locks were as fine grained as you ever needed, constraints should be handled by the client application, and that stored procedures existed because other vendors couldn't write fast SQL parsers (thus showing a failure to understand the difference between declarative and procedural programming). Now, demands from their larger customers, plus products from people like Heikki, appear to be forcing MySQL AB forwards towards ACID compliance. So, no, I do not trust MySQL AB to implement these things quickly and effectively.
If you read their TODOs, you can see most of this stuff in targeted for 4.1. That's 2 major revs away. What will Postgres be doing in that time?
BTW, your web server is serving 403s. You should probably check on that.
Historically, MySQL has been faster, and had *PROPER* support for BLOBs. Not that crappy 8k/row limit PostgresSQL has. (had?)
First. Postgres used to have a row limit which could be set at compile time from 4k to 32k. The default was 8k. This limitation was removed in 7.1, when they implemented chaining rows across multiple blocks in an efficient manner. You should note that InnoDB currently has an 8k limit....
Second. Define 'proper support for BLOBs'. Stored in-row? Dumbass. Imagine trying the DB trying to scan a table with variable size up-to-2-gig BLOBs in the way. Most every DB stores BLOBs in a seperate area from the rows, and I think if you look more closely, you'll see MySQL does the same.
Lastly, Netscape's TEXTAREA has a size limit of, IIRC, 32000 characters. If you use lempel-ziv compression (zip) you'll get that under 8k. As it happens, Postgres has a type, lztext, that does the compression transparently. So complaining about Postgres 7.0 for web work shows major lack of savvy.
If you'd expand your acronyms, you'd see that 'BLOB' is binary!. Text is not. You're talking about LOBs, and you still don't need to use them. Save them for when you need to store several megs text, which you'll find most people don't type very often, nor very quickly.
The only props I can give here is that MySQL's parser seems much more willing when working with BLOBs. At what size input does its parser break?
Slashdot Mirror
This is exactly like Intel pumping up Mhz for marketing purposes rather than actual engineering. High latency, on servers, sucks ass. RDRAM? No thanks.
"If we admire the Net, should not a burden of proof fall on those who would change the basic assumptions that brought it about in the first place?"
I was going to just write off the article as SoCal tissue, but my eye was caught at the end by this:
*BOGGLE* Oh indeed; lets have those Harvard MBAs configuring Cisco 8100s, shall we? This is a former chair of IANA! He should know better!I think this is just a fundamental difference -- myself and my friends hate the idea of a big company telling us what games we're going to like. Hate it so much, in fact, we import odd stuff from Japan when SCEA, normally taking all comers, doesn't think it'll sell. Recent acquisition: 'Vib Ribbon', the only 1.5 dimensional game in existence. Sad thing is, Sega was the same way..
We seem to be discussing different things. Let me restate/rebut: Nintendo has strong in-house teams because they couldn't trust exterior developers, whereas Sony has, well, jack shit in-house and hordes of good people outside deploying on their hardware. It's also historic: Nintendo is a games company that builds hardware to sell games; Sony is a hardware company that licenses games to sell hardware. Sort of a MS/Sun duality. And I'm happy with the latter. Bah. Hardcore for what? What makes you think what you find cool is what the rest of us find cool? I'm just going to guess that we're the same age (mid-20's); we probably have played many of the same games; we can both agree that Nintendo is powered by Shigeru Miyamoto, a designer as determined and unstoppable as a mecha come to destroy Tokyo. Even so, when was the last time you waited in a line all night? That kid stuff is over -- our generation of gamers now have days jobs and don't mind picking up a game a few days late. This kind of messaniac fervor is passing.
What's ElBo? Must be an East Coast thing...
OTOH, I bought the PSX for the explicit use of FF7. Money well spent, I should point out (break it down into $ per hour of fun - it's enlightening). I'll buy a PS2, sight unseen, to play FFX. Although I want to play the new Metroid, I know that the only good games will be the ones done by Miyamoto - there will be, what, 4 over the life of the GameCube? (metroid, 2 zeldas, etc?) Not worth it. If Sega's Smilebit team targets it, yeah, that would rule, but they're already targeting PS2, mainly because it's open to all comers.Anyway, this little discussion just proves my point -- that if two fanatical video game fans with similar backgrounds can't come to agreement on what's important in a game, a big dumb company with a list of trademarks, sales spreadsheets, axes to grind, and hordes of customers spread across many market segments has almost no chance to be able to define, let alone regulate, Quality.
This is much more like real-time programming than more conventional desktop coding, and why consoles are harder to work with than PCs. I do not think this problem will magically go away.
How has Sony's quantity over quality creed hurt them? Everyone jumped ship from Nintendo to PSX due to Nintendo's censorship, incredibly restrictive licensing system, and bad technical decisions. Sony's more open licensing system gave them the crown for the last half of the 90's. Sony's one attempt at 'quality control' was trying to restrict RPGs in a misguided attempt at branding the PSX as a action/sport console -- an attempt that hurt everyone involved, none more than Sony. They learned their lesson after seeing which segment sold the most and had the most repeat customers; and now on the PS2 they push the RPGs as the system mainstay.
Games are still a new field, with development very analogous to start-ups. A big dumb company wielding creative control is, and always has been, a dramatic failure. Sony, a big dumb company if ever there was one, has gained a great deal by keeping their hands off their developers.
Quantity, not quality, becuase you can't predict quality.
Very sad story."
p.s.: the best was certainly the 'spring of drowned octopus' -- still trying to figure out how that works.
We may need more than one station.
Asking to speak to a lawyer is the key thing; as it forces their hand to either book charges or else let you go. If they book you, there are then more stringent requirements on evidence, what they can ask you, rules of intimidation, lawyer present, etc; none of which are imposed when you aren't formally arrested. So 'detainment' is a great boon for cops, as it means every word you say can be used to build a case against you, since it's just normal conversation in the eyes of the law.
I somehow doubt there will be a "You are traffic. Take a bike or carpool." message.
Yes, these FrontBase guys are active on comp.database.*, and they're certainly easier to talk to than Oracle developers. For both of those I am glad. But the topic of the 'Ask Slashdot' is something more powerful than Postgres, and FrontBase really isn't even up to that yet, let alone in the range of Oracle and DB2. Stop spamming; read the article; stay on topic.
Americans, South Koreans, and some Europeans play PC games. Everyone else plays console games. If you look closely, the internet stopped being the killer app for PCs a couple years ago, leaving games as the prime mover for faster boxes.
Landline Net access is inhibited by NTT. Thus, Japan has the most advanced cell network in the world, and you can get all manner of net services over DoCoMo. Speaking of innovation, DoCoMo's profit sharing system has created a whole cottage industry of service innovation. To say that software development in Japan is dead is blatantly false.
There's another reason American software companies are having trouble selling in Japan, and it's not just their bad EUC/JIS/Unicode systems. Most US firms have adopted a 'reave and plunder' approach to Japan... Oracle, for instance, charges three times as much for their products as they do here.
Sorry, Economist, but you're shilling. Goodbye, respect.
"environmentalists are ... nihilists" environmentalists are the furthest thing from nihilists. We believe that mankind has a future, and the best way to assure that is to moderate our consumption to stay in line with our ability to produce. We're working to provide for our children. Who are you working for?
"Environmentalism is *not* a science." Of course it isn't. It is a political movement, backed by *everything* science has told us over the last 50 years about sustaining humanity's future.
I won't even get into your absurd attempts at psychoanalysis. Please put down the Ayn Rand (a religion if ever there was one) and grow up.
modinfo
whatever kind of package management tools you run on your system.
Logging info you would look for in dmesg:
Stop taking shit out of context. You know the poster didn't mean one mythical '.config' file.
Most of these systems offer a LOB type that stores the actual LOB in its own file on the filesystem, so you do have that option. The problems with LOB-on-filesystem is that the LOBs are then exposed to, well, any luser on the system, some of whom mistype 'rm' arguments. Besides not being under the control of the DB's security, LOB-on-fs limits locking to standard semaphores (no MVCC), adds a layer of syscalls, complicates caching, and the DB can't do very intelligent reads like it can on its own tablespaces.
Really the only reason to use LOB-on-fs is if you need to access those same objects without going through the DB, like if you want to edit a 500M LOB in emacs. Interestingly, Oracle runs a SMB server in 8i that can expose everything as an apparent filesystem, so even that reason is a bit flimsy....
As an aside, when DEC was bought by Compaq Intel bought DEC's StrongARM and new england foundry. Most of the employees left. From what I hear a goodly chunk of the StrongARM folks joined Cadence and other design firms. I don't know about the rest.
I think there's historical evidence that the only thing Intel is buying here is patents.
There's an interesting story behind HP end-of-lifeing PA-RISC and SGI eol'ing IRIX; I posted it last night on the previous Alpha story, but maybe it deserves restating:
HP teamed with Intel b/c the CEO-for-hire they had at the time (Rick Belluzio) tried to get the company to drop PA-RISC and HP-UX and move everything to Intel and WinNT (because that's the future!). I wonder in which class in biz school do they tell you to just drop 2 decades of engineering focus and end-of-life all your products at once. The predictable occured: all their server customers went to Sun, who was busy sucking up every internet customer around. Their 'high-end NT workstations' were massively undercut by, well, every PC clone maker in the world. Moral: value your uniqueness. The board managed to fire him in time to reverse some of the damage, but HP was been burned during the fastest growth period for Unix servers in years (possibly ever). They didn't need to before, but now they really need to honor those contracts with Intel.
As an aside, this same CEO-for-hire did the same manuver at SGI (end-of-life MIPS & IRIX, sell WinTel), with the same consequences. He's now working at Microsoft. It's a facinating biz study -- every commercial Unix vendor who partnered with MS & Intel was badly damaged (DEC was destroyed). Sun, who fought MS tooth and nail, thrived. Perhaps it's naive to follow behind WinTel...
No, we won't tell you where.
So I can see that these types 'know' business real well. At least the small subset that they believe is the entirety of biz, and all hell is lose when the map doesn't match the terrain. If there's a moral here, maybe it's "never hire CEOs from outside".
BTW - UT Austin victim right here! Physics department, 99. What's your damage?
create table mean_bean_machine (
id integer primary key,
description varchar(7000),
bean_bag blob
)
In schema definition and SQL parsing, the blob is a normal column. To the database internals, the blob is not stored with the rest of the table, instead a location reference is stored with the normal data that points to a less organized area where LOBs are stored. For simplicity's sake, envision the tabular data as a linear series of 8k disk blocks.(or 4k, or 2k, or whatever). Algorithms scanning the tables can quickly move down this line of blocks, because of the regularity in layout. Such algorithms could not do the same if huge chunks of LOB were in the way, so they are stored elsewhere on disk. Every RDBMS does this, the only differences are the specifics.
What you're missing is that none of this matters to the user -- this is all internal to the database's storage structure, and the great win with the relational model is the separation of physical storage and logical design.
To get back to your specifics, 1) you shouldn't have to have a seperate table for LOBs, 2) Postgres handles LOBs just fine, and 3) you should be wary of streaming BLOBs out of MySQL -- I'm not sure how that goes over given MySQL's table-level locks. I'd suggest ripping the BLOB out quick and caching it in RAM.
HTH
I just don't see any need for doom-and-gloom here.
As an aside, this same CEO-for-hire did the same manuver at SGI (end-of-life MIPS & IRIX, sell WinTel), with the same consequences. He's now working at Microsoft. It's a facinating biz study -- every commercial Unix vendor who partnered with MS & Intel was badly damaged (DEC was destroyed). Sun, who fought MS tooth and nail, thrived. Perhaps it's naive to follow behind WinTel...
Your link leads to a Beowulf integrator. If you work there, you understand the current limits of parallel processing, and the premium companies will pay for really high end equipment. Besides, the hardware is usually the cheapest thing in an organization. And what the hell are you talking about dumping Alpha when IA64 hits the street? Compaq's Alpha business is an 8 billion per year market. Most of those are big corporate customers who won't change platforms for 5 years minimum. Assuming they'll change over this fall presupposes the existence of your magical pet monkey.
After adding all the above, MySQL will need some serious (read: light-years of) development on their query analyzer. MySQL does well on simple selects but performs notably slow on complex queries. Which is hardly a surprise given their assumptions, design goals, and evangelism.
InnoDB needs more testing, and it needs to continue to grow. I have full faith in Heikki Turri's skill.
I cannot say the same for the other MySQL AB developers. They have spent the last few years stonewalling efforts to turn MySQL into a real DB. They proclaimed that transactions were bad, table locks were as fine grained as you ever needed, constraints should be handled by the client application, and that stored procedures existed because other vendors couldn't write fast SQL parsers (thus showing a failure to understand the difference between declarative and procedural programming). Now, demands from their larger customers, plus products from people like Heikki, appear to be forcing MySQL AB forwards towards ACID compliance. So, no, I do not trust MySQL AB to implement these things quickly and effectively.
If you read their TODOs, you can see most of this stuff in targeted for 4.1. That's 2 major revs away. What will Postgres be doing in that time?
BTW, your web server is serving 403s. You should probably check on that.
Second. Define 'proper support for BLOBs'. Stored in-row? Dumbass. Imagine trying the DB trying to scan a table with variable size up-to-2-gig BLOBs in the way. Most every DB stores BLOBs in a seperate area from the rows, and I think if you look more closely, you'll see MySQL does the same.
Lastly, Netscape's TEXTAREA has a size limit of, IIRC, 32000 characters. If you use lempel-ziv compression (zip) you'll get that under 8k. As it happens, Postgres has a type, lztext, that does the compression transparently. So complaining about Postgres 7.0 for web work shows major lack of savvy.
If you'd expand your acronyms, you'd see that 'BLOB' is binary!. Text is not. You're talking about LOBs, and you still don't need to use them. Save them for when you need to store several megs text, which you'll find most people don't type very often, nor very quickly.
The only props I can give here is that MySQL's parser seems much more willing when working with BLOBs. At what size input does its parser break?