Couldn't agree more. I thought this was a thoughtful, well-organized essay, and definitely merits a discussion on Slashdot (which discussion, of course, is shaping up to be neither thoughtful nor well-organized). I also think a base "Data Type Manager" is an interesting idea that merits some thought and experimentation, and to the extent that this treatise and discussion encourage that, it's a great thing.
One of the subtle ideas this (Activity) Type Manager approach brings up is the difference between task-based and activity-based software. Back when I was on the KDE usability list, we did a lot of talking (and a lot less acting) on the subject of task-based start menus, control panels, and applications, in an attempt to get away from content-based ones. You very quickly run into the problem that there are a lot of tasks, and some of them are used in a variety of ways. But an activity ("deal with music using your computer") is big enough and happily amorphous enough that it just might bridge that gap. Another nice idea about the Activity Type Manager is that it can take on the job of figuring what metadata is important for that activity (and associated tasks) and deal with capturing and organizing that metadata.
There are some big drawbacks to this approach, namely that it requires grouping things into categories again ("activities"), and that produces a whole new set of cross-activity aspects that people have to work with, which vastly increases the complexity of the software.
Nonetheless, it's an interesting idea and worthy of discussion.
Good point. Not only is the supercomputer market a small one, as many others have pointed out, it's not a broadly lucrative space - how many scientific operations have spare cash lying around.
However, below the weather prediction, protein folding and quantum chemistry type problems is a whole tier of computation on less-than-super computers. Nearly every scientific department at a university has a couple of these: 16-100 proc little gremlins that run people's models and analysis code. Not to mention automakers (computational fluid dynamics anyone) and financial modelers. And those gremlins ain't cheap either. MS is likely trying to establish their bonafides at the top of the market so they can sell to this next tier. If you run Windows for your domain and email services, your admins are more likely to be happy with the idea that you can buy a "proven" clustering solution that lets you use a GUI to add more boxen to a cluster. Given how far Linux has penetrated in the financial industry, this may be a losing gambit there, but it's my best guess of what the gambit is.
Very, very well put. Battery-powered vehicles have another couple big reason to support them beyond emissions.
First, electric engines have a much higher limiting efficiency than combustion engines, at almost any power output. Simply put, electricity is easier to turn into mechanical motion than the chemical energy in hydrocarbons. That means power-hungry drivers can get the power they love at lower energy cost.
Second, by using gas for cars, we are committing ourselves to running two parallel and totally non-interoperable energy distribution infrastructures, which in itself is massively wasteful and polluting, quite aside from the polluting output of the hydrocarbon energy. At least when it comes to motion-making (the converse of #1 is that electricity to heat is a very poor conversion), we should be pushing for a combined distribution system, with modular inputs and outputs. This compatible-architecture gives you the same kinds of benefits as the Internet: open standards for energy are good just like in software.
Given that a perfectly functional electricty infrastructure already exists, getting power to most commuter cars is pretty straightforward: some digitally lockable power cords at your parking garage or meter that can deal with charging for power. Or some system of exchanging drained batteries for charged ones. None of which is that hard, particularly if the gov't chips in some $$$ to get the ball rolling.
Third, the most promising portable energy solutions all point towards electric engines: fuel cells, hydrogen, etc. So we should be getting as many kinks as possible worked out of electric car engines, including performance, disposal, fabrication supply chain, etc, as they are the future.
The fact that an implementable technology like batteries has been completely shunted aside in favor of vapordrive is indeed infuriating.
Doomed is probably a bit strong, as humans are awfully ingenious animals. We certainly seem likely to be suffering some nasty consequences for a long time.
Another reason to tax oil that deserves discussion is price stability. Of the economically damaging effects of peak oil, price volatility is likely to be far worse than rising prices themselves. People can adapt to predictably rising prices, but unexpected swings make for a very difficult business climate (and/or expense if you have to insure yourself against them). If the government intervenes in the energy market via strong taxation, not only does this allow the country to invest in forward-looking energy sources, it allows the government an easy way to stabilize prices. Wahabbis blow up a plurality of the wells in Saudi Arabia? Temporarily halve gas taxes to keep oil from tripling in price. Oil glut causes gas prices to plunge? Add a little tax to keep demand from spiking. I'm sure the libertarians will bitch and moan about the government intervening in a market, but it's precisely this kind of intervention that got the US and the world out of the Great Depression. Peak oil could easily be far more catastrophic than the Great Depression, and can be avoided or minimized by a sensible energy policy now.
I'd be curious to know how people think Singularity compares to the Hurd. IIRC, the Hurd looks at a lot of these ideas and takes the opposite tack from Singularity: rather than run type-checked, managed processes in kernel space, push everything out to userspace, and broker interaction with the kernel via various server processes. Has anyone done a comparison of the merits of these two approaches?
Yeah, the mags really are making a mountain out of a molehill here. And there's really very little here that couldn't have been done 3-5 years ago - browsers had the same javascript support, SOAP existed, and people knew you could make a request from within Javascript.
Am I missing something? I've always thought that was part of why people didn't do this before - amount of coding needed to implement a simple app is vastly more than with something like.NET or QT. Is that not true? Are there AJAX development kits, or do you have to build the javascript (and corresponding web services) for every little behavior by hand? And what about the problems of implementing cross-browser Java-script? Has that been simplified?
No, he doesn't. That's exactly the point of all of this. If he knew what sells, he wouldn't have been blindsided by the success of Google's business model, and start yet another round of frantic catch-up to superior emergent technology from another company.
I beg to differ. Nobody can know everything, and the fact that potential opportunities passed by one group of people is just how the world works. I think about this so I can't think about that. Part of being a great businessman is accepting that you missed the party at its very outset, and get in early enough. Gates and Microsoft have been astoundingly successful at that.
Microsoft in general has a very hard time understand that people may or may not want to do it the way they've laid out, and that's where OSS really has made strides. Microsoft does have a strong developer community, but it all focuses around "this is how this feature X and how you use it", rather than OSS': "this is a good way to do Y", or even more "what if we did Y instead of X?".
An MS employee friend of mine who's a longtime OpenBSD user did a report on Linux and BSD for some VP's a few years ago, and all they questions he heard were "does it do X,Y, or Z?" "Does it do them seamlessly?", and they dismissed it. He hard a very hard time conveying that many developers like the ability to open the hood and see what's going on.
Microsoft has missed the Internet once before, but it was lucky in two ways: 1) neither consumers nor infrastructure were ready for a full-scale consumer Internet 2) adopting the internet merely meant making their software use it compellingly. They needed to deliver a better browsing experience with the OS, as well as crush the browser as a standalone application platform and with IE5 they did both. It certainly involved a lot of dirty tricks, but it was also a hell of a business feat.
This time, however, the OSS developer community has made people think hard about open standards, as well as providing commodity tools to build services with. Leaving aside the question of whether the GoogleFarm could run well on Windows, why would you pay $200 per copy and relinquish your rights to modify. As a consumer I don't care what OS a service provider runs, so they're likely to pick the cheapest one that works.
Microsoft is too big to disappear, and people will want desktop computer functionality for a long time to come. They will certainly make many more billions. And I wouldn't be surprised if MS engineered an IBM-like transformation to a more mature company in a particular space. But it's hard to see Microsoft maintaining a position of dominance over the next 5 years.
I don't think I quite agree with that. You don't have to understand all the stuff that goes on with indexing, query planning, system tables, etc, to get going with Postgres. But when you need it, it's open, extensible and documented. I say that because I think Postgres is doing an increasingly good job making the software work for people of various levels of expertise.
No flames from me. I happen to agree with you, especially since Postgres is now well-position to be desktop database engine of choice, thanks to its small footprint, unrestrictive license, and ease of manageability. So a simplified installer and some wizards (make a user), would make it a great database to get started doing Access-like things.
Good point. You have more substance to what was an intuitive arugment from me - that Google only has that kind of profit margin because they are so dominant in search. What would happen if they were the size of Altavista? After all, their ads sell for pennies per thousand page views. In other words, what's their margin per customer or per sale?
I think that's the problem many OSS business face - the margin on a sale is constrained, often because they are competing in a commodity or downmarket space.
Google is both a good example and a bad example. It's a good example because it operates on very narrow margins, and business models around open source software almost always require very narrow margins: cost of production goes down, but a large fraction of users do not become paying customers (I believe it's 2% for most OSS companies).
It's a bad example because it was technologically advanced enough to dominate an emerging space, so despite its narrow margins, it can dominate the space. Most OSS companies I've seen are not in that enviable position - rather they commoditizing an existing space, which doesn't give them the big bucks. Don't get me wrong, I believe there's good money to be made in commodity spaces, but I doubt you can rocket to prominence there.
So I think a number of VC's are going to be unpleasantly suprised when their 3 year in-and-out strategy turns into a 10 year plan for modest growth.
Open source may be garnerning more investment than it merits, but it's still a very small fraction of the software development market, much less the overall enterprise IT sales and service markets. So if OSS investment overheats, it will just lead to VC's being wary of investing in OSS companies for a while. Given that OSS is more a process than a particular sector or a business model, this will have limited effect on the overall march of open source, and especially open standards. In concrete terms, it would suck for JasperSoft to lose funding because hype around OSS' profit potential turns out to be overrated, but it wouldn't really stop people like Sun or IBM from moving towards open source, nor will it stop the commoditization of a variety of products by open source equivalents.
I respect your right to believe what you like about ID, both as a thinking person, and of course from a freedom of person perspective.
But I have two reasons to find Intelligent Design troubling. First its primary use has been not to try to "falsify" or find and resolve problems in various theories of human origins, but rather to posit that a particular explanation of human origins (i.e., a biblical one) is the right one and also that it cannot be proven or disproven by science, thereby calling biblical explanations of human origins scientific without subjecting them to empirical tests conducted in the manner of scientific inquiry. This is why the scientific community finds ID as science troubling: it is used to postulate the veracity of an untestable theory.
If Intelligent Design is in fact science, what experiment could possibly be conducted that could disprove the notion that some parts of biology are irreducibly complex and could not have occurred naturally? In other words, what kind of evidence would convince you that ID was incorrect? Philosophically there are no settled truths in science, only theories that nobody has been able to prove wrong. In practice of course, things that have been verified extensively are often taken as settled truth, but it is scientifically valid to ask questions of the laws of gravitation, or cell division, even algebra. Evolutionary biology and science in general is built on these kinds of questions - by attempting to find evidence that cannot be adequately explained by a particular theory, and then attempting to formulate a theory that incorporates the new evidence and the old ones.
I don't happen to have an opinion on whether a Creator started the universe, or interceded in some way, and of course I think that's an interesting question. Inasmuch as you're saying that the direction of evolution has leaned towards but not verified the notion that a Creator could not have cause human life, I think there's a legitimate argument to be had about what science has or has not spoken to.
It's also fine to say that you believe that science cannot describe the origins of life of the universe (or more importantly, the question of why we came to be, and what it means), but in that case, ID is again not a scientific theory.
We have optical hard disks, and they are a hella of a lot slower than magnetic ones. The optics we're talking about here are for moving the signal around the machine (and over the network) after it's been read from the media.
My guess is that there are still some nasty snags awaiting even making a serious optical router, much less producing it commercially. I'm betting more on 2012 than 2007. Hell, even LongVista won't be out by 2007.
The OCA [Open Content Alliance] will seek to digitize all public domain works, but only copyright material for which they gain explicit consent from the publisher. Made up of Google competitors Yahoo! and the Microsoft Network (MSN)
Open is the key word here, as in open standards and terms for indexing and searching. Let's look at the trends. E-book readers, while not widespread, are now seriously viable, thanks to the imminent release of usable electronic paper, as well as iPods and laptops. Consumers are starting to get interested in the idea getting book-length media via the Internet. If Google successfully builds a monopoly on searching for books, they will be able to shunt those searches off to whoever the (re)sellers of the electronic books are, and obligate them to obey their terms. Publishers lose the main value-add part of their services, marketing and distribution, and become editors for hire. Doubly so in a Google Printed world because Google's main goal is to do control marketing and distribution (and profit from ad sales and referrals as the transaction flows by).
This sounds eerily reminiscent of every/.er's favorite company and their media player, or of the control (in the US) cellphone service providers exert over cellphone hardware. Remember when Microsoft was trying control SOAP/WSDL/DTD, and how everyone insisted they should work with standards bodies? Time to insist on that again, or we'll be buying only Google brand books, and end up lamer/. trolls writing about how 6006L3 S4X0R for the next decade.
1> The application should be sandboxed in as many ways as possible. Microsoft has always been convinced that a global digital signature verification system is just over the horizon (preferrably controlled by them of course), and that will resolve the issue of verifying that an application is secure (AFAICT whole ActiveX debacle was built around the assumption that enforceable digital signatures would just come around and secure the nasty holes ActiveX opened up). But if the world can't even agree on how to run a domain name registry, I have little hope that we'll get a universal application verifying authority soon. So sandboxing is key for the foreseeable future.
2> Dependency hell. The best thing I've heard in ages is DragonflyBSD's idea to have a masking system that lets programs overlay libraries temporarily, until the dependencies can be resolved to a single library, I hope by a dependency-walker that could run later. Not only would this bypass RPM hell, but I imagine it could also be used to keep system libraries more secure.
3> RIA's are a big deal. Ok, they're not quite here yet but within 2 years they'll be pretty common. The distinction between an application in a browser, on in a terminal client like Citrix or NX and a local one should be minimized, so installing and running feels nearly the same for all of them.
The stuff about controls (widgets) is just going back to Microsoft's basic worldview. For the last 10 years, Microsoft's applications have generally had outstanding control sets. Java, for example, does not. It's true that that makes their applications comfortable to use, but the way MS controls are built has always kind of pinned you into a client-heavy architectures, though I think.NET makes some good progress on this. Maybe I'm wrong, but I'm not convinced it's that big a deal going forward.
I can't speak for regulations in the medical industry, but I have had to certify point-of-sale applications that involve communication with credit/debit-card authorisation systems. Trust me--it does not matter if you use Oracle or MSSQL or PostgreSQL as your database engine--you still have to pay the testing fees and do the battery of verification tests for your application before they permit you to connect to the live system.
To be honest, I'm just starting down that road on the medical side of things, so I defer to your experience. However, I'm pretty sure that you can be asked to validate not just the performance but the design of components of software, like libraries and databases. Which can be harder in a distributed project like Postgres, since there's no clear repository of design documents, nor a particular build control process. Perhaps someone like Pervasive will take to providing the backing for this kind of validation...certainly the kind of money involved precludes a volunteer community from undertaking it. In any event, I'm all for people "certifying" Postgres in whatever way they like - anything that raises its profile is a good step.
I tried using Firebird for a 6-client project 2 years ago, because Postgres hadn't come out on Windows at that time. I found the documentation pretty weak overall, the admin tools were very immature, and the completeness of the features was overhyped. For example, I never could use updatable views with Access (they work just fine with Postgres), and never could nail down acknowledgement from anyone that it was a bug. Maybe things have gotten better since then, but I was totally underwhelmed. I switched to Postgres and have never looked back.
in the case of communications services that is why we impose government regulation, which in turn creates a whole new set of middlemen but this time with guns.
..and the government at least has a charter to represent the public interest. When regulation is done properly, it means the regulators strike a balance between consumer interests (public as consumer) and business interests (public as worker and investor). The problem is that for 20 years there have been no serious efforts to make any forward reaching regulation, under the various arguments that "regulation always makes things worse", and "the government can't keep up with the market". There's some truth to these criticisms, but AFAICT the main problem has always been that we continue to allow vertical integration between a competitive market (carrier services) and natural-monopoly public infrastructure (phone lines and bandwidth). The minute we separate them, we can deregulate the carrier market all we like, and it will promptly commoditize. Telephone lines can be kept either a a government-provided service (which would probably make their quality work at at about the level of roads), or go back to utility-style control of the maintenance providers.
This CEO says he "owns" the pipes. Fine, let's get the government all the way out of this venture. I want a reckoning of much money local, state, federal governments have put into the building and maintenance of those pipes. And if SBC's going to "own" them, they'd better cut those governments some big ass checks to compensate them for their investments, and the government can plow that into making the communications market competitive again. Otherwise, I hope the state AG's start looking hard in SBC's direction...
Oracle also has a number of certifications that make it very attractive in certain industries like defense, healthcare, etc. I work in the medical field, and a free 4GB-max "validated" Oracle database is a huge boon to a wide variety of medical ISV's. Given that's it free to redistribute, the OSS/proprietary thing kind of takes a backseat for a lot of these ISVs.
By the same token, I also don't think it's going to drive sales in the way they think it will. Databases are slowly but surely going commodity, at least at the lower end of the market, and this merely reinforces that trend. And along with that, there's an increasingly robust set of tools to obviate the differences between these database for most uses that don't demand extreme peformance, from Hibernate and ORM packages to ADODB and other database-independence layers in PHP to.NET's layered data architecture.
As a Postgres user, I'm hopeful that Sun's proclaimed interest in Postgres will result in this kind of "validation". However, given Sun's reputedly somewhat lackadaisical commitment to staffing OOo, I'm not holding my breath. With Postgres' extensibility and extremely high-caliber core developer base, I think a strong commitment to validation by Sun could make it a real contender in the medium enterprise space. Validate it, clean up a few features (notably auto-vacuum and passable auto-tuning, maybe some multi-master replication), throw in a simple deployment for ORM or database indirection, and you've effectively moved that commoditization up one layer from the small website developer level.
In the long run, I don't see how this gets Oracle out of the need to transition its core revenue off of its database licenses.
That's true that in order to be semantically power it would need to be complex. I was thinking that it wouldn't need to run for regular config file reads, tho - maybe some kind of file alteration monitor could inform the syntax-checker when files were changed. Anyway, that's enough attention to an idle thought... Everything is indeed a tradeoff.
The registry's idea of a centralized (and as you point out, distributable/replicable) database style solution for configuration is in itself not a bad idea. However, configuration is one of those situations that begs for tree-like information, rather than flat relation information - hence the slow but steady migration of almost every text configuration file towards XML or a variant. So if it's powered by a DB thoroughly comfortable with trees, that's great. And it may finally be getting there.
But there's still very little that's as robust as a text file and a filesystem. I think a happy medium would be something that enforced some kind of syntax on all config files, and provided a single interface to them when you wanted it, but still allowed them to work basically as text files.
IIRC there was an article not too long ago that explained that the main difference between Office and OO is that Office makes extensive use of lazy loading, while OO essentially hammers through loading every library it may need, which not only thrashes the disk once on initial load, but again as you (likely) swap out memory pages. My recollection was that this lack of lazy loading had something to do with cross-platform compiling and linking issues, as well as MS having extensive resources to put into optimizing Office loading that OO did not. My understanding is that Sun hasn't exactly dumped developer time into OO, either, and I believe the focus of this release was compatibility and.
Typically people solve this problem by preloading a bunch of the relevant libraries at startup, a strategy both MS and OO attempt to employ (viz OfficeStartup and OO QuickStarter). I used to detest that, but if I had 1 or 2GB or RAM and wanted to rely on OO, I might not find it so bad. I think an interesting addition to this comparison would be to see how OO fared with QuickStarter enabled, and what drain that placed on the rest of the system. Likewise disabling the JVM loading.
The wonderful thing about computer science/software engineering (and the terrible thing about software patents) is that everyone can borrow ideas from one another to advance the environment as a whole.
Spot on. THis is totally true of ClearType: there were some font-improvement and sub-pixel hinting idea kicking around out there, and MS brought them together and built a nice usable engine for them. Kudos..NET is an interesting amalgam. It definitely borrows very heavily from Java (C# is so Java-esque it's absurd), and combines a basic idea with a very forward-looking one: obviously MS needed a way to bind to its libraries (MFC/Win32 API) from a variety of languages (like the KDE language bindings), but it wanted to make all them work within a JIT compiler system. Thus the idea of "hosted" languages all running within a common runtime, which in my mind is a very forward-thinking approach. It will be interesting to see if this PHP joining Eclipse makes any push in that same direction.
I also agree with you that Microsoft has built some great software in its day. Excel continues to set the standard for spreadsheets, and underpins a huge amount of business analytics. When IE5 was released it broke new ground for browser usability. They've also built some shite, and the fact that they're so insistent on the "my way or the highway" approach makes working with their software totally aggravating.
MS really does have a lot of catch-up to do in the hosting environment. My experiences dealing with Windows as a hosting server have been miserable - it's almost impossible to tell what the current configuration of some component unless it's exposed through some control panel type page, and even if you do figure out how it's configured you generally have to have an Administrator make some GUI change. That's not to suggested that *nix hosted doesn't have its own frustrations, but you can usually figure out more less what's going on. It looks like Monad represents a new direction in that sense, but they really need to do something to expose the registry via command line to make it more transparent, or better yet, ditch the registry. I'll bet the registry accounts for more than half of the unrecoverable Windows corruptions I've seen, and it's a performance bottleneck to boot. We'll see if they can pull off that kind of change.
Slashdot Mirror
Couldn't agree more. I thought this was a thoughtful, well-organized essay, and definitely merits a discussion on Slashdot (which discussion, of course, is shaping up to be neither thoughtful nor well-organized). I also think a base "Data Type Manager" is an interesting idea that merits some thought and experimentation, and to the extent that this treatise and discussion encourage that, it's a great thing.
One of the subtle ideas this (Activity) Type Manager approach brings up is the difference between task-based and activity-based software. Back when I was on the KDE usability list, we did a lot of talking (and a lot less acting) on the subject of task-based start menus, control panels, and applications, in an attempt to get away from content-based ones. You very quickly run into the problem that there are a lot of tasks, and some of them are used in a variety of ways. But an activity ("deal with music using your computer") is big enough and happily amorphous enough that it just might bridge that gap. Another nice idea about the Activity Type Manager is that it can take on the job of figuring what metadata is important for that activity (and associated tasks) and deal with capturing and organizing that metadata.
There are some big drawbacks to this approach, namely that it requires grouping things into categories again ("activities"), and that produces a whole new set of cross-activity aspects that people have to work with, which vastly increases the complexity of the software.
Nonetheless, it's an interesting idea and worthy of discussion.
Good point. Not only is the supercomputer market a small one, as many others have pointed out, it's not a broadly lucrative space - how many scientific operations have spare cash lying around.
However, below the weather prediction, protein folding and quantum chemistry type problems is a whole tier of computation on less-than-super computers. Nearly every scientific department at a university has a couple of these: 16-100 proc little gremlins that run people's models and analysis code. Not to mention automakers (computational fluid dynamics anyone) and financial modelers. And those gremlins ain't cheap either. MS is likely trying to establish their bonafides at the top of the market so they can sell to this next tier. If you run Windows for your domain and email services, your admins are more likely to be happy with the idea that you can buy a "proven" clustering solution that lets you use a GUI to add more boxen to a cluster. Given how far Linux has penetrated in the financial industry, this may be a losing gambit there, but it's my best guess of what the gambit is.
Very, very well put. Battery-powered vehicles have another couple big reason to support them beyond emissions.
First, electric engines have a much higher limiting efficiency than combustion engines, at almost any power output. Simply put, electricity is easier to turn into mechanical motion than the chemical energy in hydrocarbons. That means power-hungry drivers can get the power they love at lower energy cost.
Second, by using gas for cars, we are committing ourselves to running two parallel and totally non-interoperable energy distribution infrastructures, which in itself is massively wasteful and polluting, quite aside from the polluting output of the hydrocarbon energy. At least when it comes to motion-making (the converse of #1 is that electricity to heat is a very poor conversion), we should be pushing for a combined distribution system, with modular inputs and outputs. This compatible-architecture gives you the same kinds of benefits as the Internet: open standards for energy are good just like in software.
Given that a perfectly functional electricty infrastructure already exists, getting power to most commuter cars is pretty straightforward: some digitally lockable power cords at your parking garage or meter that can deal with charging for power. Or some system of exchanging drained batteries for charged ones. None of which is that hard, particularly if the gov't chips in some $$$ to get the ball rolling.
Third, the most promising portable energy solutions all point towards electric engines: fuel cells, hydrogen, etc. So we should be getting as many kinks as possible worked out of electric car engines, including performance, disposal, fabrication supply chain, etc, as they are the future.
The fact that an implementable technology like batteries has been completely shunted aside in favor of vapordrive is indeed infuriating.
Doomed is probably a bit strong, as humans are awfully ingenious animals. We certainly seem likely to be suffering some nasty consequences for a long time.
Another reason to tax oil that deserves discussion is price stability. Of the economically damaging effects of peak oil, price volatility is likely to be far worse than rising prices themselves. People can adapt to predictably rising prices, but unexpected swings make for a very difficult business climate (and/or expense if you have to insure yourself against them). If the government intervenes in the energy market via strong taxation, not only does this allow the country to invest in forward-looking energy sources, it allows the government an easy way to stabilize prices. Wahabbis blow up a plurality of the wells in Saudi Arabia? Temporarily halve gas taxes to keep oil from tripling in price. Oil glut causes gas prices to plunge? Add a little tax to keep demand from spiking. I'm sure the libertarians will bitch and moan about the government intervening in a market, but it's precisely this kind of intervention that got the US and the world out of the Great Depression. Peak oil could easily be far more catastrophic than the Great Depression, and can be avoided or minimized by a sensible energy policy now.
I'd be curious to know how people think Singularity compares to the Hurd. IIRC, the Hurd looks at a lot of these ideas and takes the opposite tack from Singularity: rather than run type-checked, managed processes in kernel space, push everything out to userspace, and broker interaction with the kernel via various server processes. Has anyone done a comparison of the merits of these two approaches?
Yeah, the mags really are making a mountain out of a molehill here. And there's really very little here that couldn't have been done 3-5 years ago - browsers had the same javascript support, SOAP existed, and people knew you could make a request from within Javascript.
.NET or QT. Is that not true? Are there AJAX development kits, or do you have to build the javascript (and corresponding web services) for every little behavior by hand? And what about the problems of implementing cross-browser Java-script? Has that been simplified?
Am I missing something? I've always thought that was part of why people didn't do this before - amount of coding needed to implement a simple app is vastly more than with something like
Microsoft in general has a very hard time understand that people may or may not want to do it the way they've laid out, and that's where OSS really has made strides. Microsoft does have a strong developer community, but it all focuses around "this is how this feature X and how you use it", rather than OSS': "this is a good way to do Y", or even more "what if we did Y instead of X?".
An MS employee friend of mine who's a longtime OpenBSD user did a report on Linux and BSD for some VP's a few years ago, and all they questions he heard were "does it do X,Y, or Z?" "Does it do them seamlessly?", and they dismissed it. He hard a very hard time conveying that many developers like the ability to open the hood and see what's going on.
Microsoft has missed the Internet once before, but it was lucky in two ways:
1) neither consumers nor infrastructure were ready for a full-scale consumer Internet
2) adopting the internet merely meant making their software use it compellingly. They needed to deliver a better browsing experience with the OS, as well as crush the browser as a standalone application platform and with IE5 they did both.
It certainly involved a lot of dirty tricks, but it was also a hell of a business feat.
This time, however, the OSS developer community has made people think hard about open standards, as well as providing commodity tools to build services with. Leaving aside the question of whether the GoogleFarm could run well on Windows, why would you pay $200 per copy and relinquish your rights to modify. As a consumer I don't care what OS a service provider runs, so they're likely to pick the cheapest one that works.
Microsoft is too big to disappear, and people will want desktop computer functionality for a long time to come. They will certainly make many more billions. And I wouldn't be surprised if MS engineered an IBM-like transformation to a more mature company in a particular space. But it's hard to see Microsoft maintaining a position of dominance over the next 5 years.
I don't think I quite agree with that. You don't have to understand all the stuff that goes on with indexing, query planning, system tables, etc, to get going with Postgres. But when you need it, it's open, extensible and documented. I say that because I think Postgres is doing an increasingly good job making the software work for people of various levels of expertise.
This is a good point, and something I was always sort bothered by. However, all the shared MySQL installations I've ever used have this same setup...
No flames from me. I happen to agree with you, especially since Postgres is now well-position to be desktop database engine of choice, thanks to its small footprint, unrestrictive license, and ease of manageability. So a simplified installer and some wizards (make a user), would make it a great database to get started doing Access-like things.
Good point. You have more substance to what was an intuitive arugment from me - that Google only has that kind of profit margin because they are so dominant in search. What would happen if they were the size of Altavista? After all, their ads sell for pennies per thousand page views. In other words, what's their margin per customer or per sale?
I think that's the problem many OSS business face - the margin on a sale is constrained, often because they are competing in a commodity or downmarket space.
Google is both a good example and a bad example. It's a good example because it operates on very narrow margins, and business models around open source software almost always require very narrow margins: cost of production goes down, but a large fraction of users do not become paying customers (I believe it's 2% for most OSS companies).
It's a bad example because it was technologically advanced enough to dominate an emerging space, so despite its narrow margins, it can dominate the space. Most OSS companies I've seen are not in that enviable position - rather they commoditizing an existing space, which doesn't give them the big bucks. Don't get me wrong, I believe there's good money to be made in commodity spaces, but I doubt you can rocket to prominence there.
So I think a number of VC's are going to be unpleasantly suprised when their 3 year in-and-out strategy turns into a 10 year plan for modest growth.
Open source may be garnerning more investment than it merits, but it's still a very small fraction of the software development market, much less the overall enterprise IT sales and service markets. So if OSS investment overheats, it will just lead to VC's being wary of investing in OSS companies for a while. Given that OSS is more a process than a particular sector or a business model, this will have limited effect on the overall march of open source, and especially open standards. In concrete terms, it would suck for JasperSoft to lose funding because hype around OSS' profit potential turns out to be overrated, but it wouldn't really stop people like Sun or IBM from moving towards open source, nor will it stop the commoditization of a variety of products by open source equivalents.
I respect your right to believe what you like about ID, both as a thinking person, and of course from a freedom of person perspective.
But I have two reasons to find Intelligent Design troubling. First its primary use has been not to try to "falsify" or find and resolve problems in various theories of human origins, but rather to posit that a particular explanation of human origins (i.e., a biblical one) is the right one and also that it cannot be proven or disproven by science, thereby calling biblical explanations of human origins scientific without subjecting them to empirical tests conducted in the manner of scientific inquiry. This is why the scientific community finds ID as science troubling: it is used to postulate the veracity of an untestable theory.
If Intelligent Design is in fact science, what experiment could possibly be conducted that could disprove the notion that some parts of biology are irreducibly complex and could not have occurred naturally? In other words, what kind of evidence would convince you that ID was incorrect? Philosophically there are no settled truths in science, only theories that nobody has been able to prove wrong. In practice of course, things that have been verified extensively are often taken as settled truth, but it is scientifically valid to ask questions of the laws of gravitation, or cell division, even algebra. Evolutionary biology and science in general is built on these kinds of questions - by attempting to find evidence that cannot be adequately explained by a particular theory, and then attempting to formulate a theory that incorporates the new evidence and the old ones.
I don't happen to have an opinion on whether a Creator started the universe, or interceded in some way, and of course I think that's an interesting question. Inasmuch as you're saying that the direction of evolution has leaned towards but not verified the notion that a Creator could not have cause human life, I think there's a legitimate argument to be had about what science has or has not spoken to.
It's also fine to say that you believe that science cannot describe the origins of life of the universe (or more importantly, the question of why we came to be, and what it means), but in that case, ID is again not a scientific theory.
We have optical hard disks, and they are a hella of a lot slower than magnetic ones. The optics we're talking about here are for moving the signal around the machine (and over the network) after it's been read from the media.
My guess is that there are still some nasty snags awaiting even making a serious optical router, much less producing it commercially. I'm betting more on 2012 than 2007. Hell, even LongVista won't be out by 2007.
This sounds eerily reminiscent of every
Yes, yes and yes. Let me add a few more points.
.NET makes some good progress on this. Maybe I'm wrong, but I'm not convinced it's that big a deal going forward.
1> The application should be sandboxed in as many ways as possible. Microsoft has always been convinced that a global digital signature verification system is just over the horizon (preferrably controlled by them of course), and that will resolve the issue of verifying that an application is secure (AFAICT whole ActiveX debacle was built around the assumption that enforceable digital signatures would just come around and secure the nasty holes ActiveX opened up). But if the world can't even agree on how to run a domain name registry, I have little hope that we'll get a universal application verifying authority soon. So sandboxing is key for the foreseeable future.
2> Dependency hell. The best thing I've heard in ages is DragonflyBSD's idea to have a masking system that lets programs overlay libraries temporarily, until the dependencies can be resolved to a single library, I hope by a dependency-walker that could run later. Not only would this bypass RPM hell, but I imagine it could also be used to keep system libraries more secure.
3> RIA's are a big deal. Ok, they're not quite here yet but within 2 years they'll be pretty common. The distinction between an application in a browser, on in a terminal client like Citrix or NX and a local one should be minimized, so installing and running feels nearly the same for all of them.
The stuff about controls (widgets) is just going back to Microsoft's basic worldview. For the last 10 years, Microsoft's applications have generally had outstanding control sets. Java, for example, does not. It's true that that makes their applications comfortable to use, but the way MS controls are built has always kind of pinned you into a client-heavy architectures, though I think
I tried using Firebird for a 6-client project 2 years ago, because Postgres hadn't come out on Windows at that time. I found the documentation pretty weak overall, the admin tools were very immature, and the completeness of the features was overhyped. For example, I never could use updatable views with Access (they work just fine with Postgres), and never could nail down acknowledgement from anyone that it was a bug. Maybe things have gotten better since then, but I was totally underwhelmed. I switched to Postgres and have never looked back.
This CEO says he "owns" the pipes. Fine, let's get the government all the way out of this venture. I want a reckoning of much money local, state, federal governments have put into the building and maintenance of those pipes. And if SBC's going to "own" them, they'd better cut those governments some big ass checks to compensate them for their investments, and the government can plow that into making the communications market competitive again. Otherwise, I hope the state AG's start looking hard in SBC's direction...
Oracle also has a number of certifications that make it very attractive in certain industries like defense, healthcare, etc. I work in the medical field, and a free 4GB-max "validated" Oracle database is a huge boon to a wide variety of medical ISV's. Given that's it free to redistribute, the OSS/proprietary thing kind of takes a backseat for a lot of these ISVs.
.NET's layered data architecture.
By the same token, I also don't think it's going to drive sales in the way they think it will. Databases are slowly but surely going commodity, at least at the lower end of the market, and this merely reinforces that trend. And along with that, there's an increasingly robust set of tools to obviate the differences between these database for most uses that don't demand extreme peformance, from Hibernate and ORM packages to ADODB and other database-independence layers in PHP to
As a Postgres user, I'm hopeful that Sun's proclaimed interest in Postgres will result in this kind of "validation". However, given Sun's reputedly somewhat lackadaisical commitment to staffing OOo, I'm not holding my breath. With Postgres' extensibility and extremely high-caliber core developer base, I think a strong commitment to validation by Sun could make it a real contender in the medium enterprise space. Validate it, clean up a few features (notably auto-vacuum and passable auto-tuning, maybe some multi-master replication), throw in a simple deployment for ORM or database indirection, and you've effectively moved that commoditization up one layer from the small website developer level.
In the long run, I don't see how this gets Oracle out of the need to transition its core revenue off of its database licenses.
That's true that in order to be semantically power it would need to be complex. I was thinking that it wouldn't need to run for regular config file reads, tho - maybe some kind of file alteration monitor could inform the syntax-checker when files were changed. Anyway, that's enough attention to an idle thought... Everything is indeed a tradeoff.
The registry's idea of a centralized (and as you point out, distributable/replicable) database style solution for configuration is in itself not a bad idea. However, configuration is one of those situations that begs for tree-like information, rather than flat relation information - hence the slow but steady migration of almost every text configuration file towards XML or a variant. So if it's powered by a DB thoroughly comfortable with trees, that's great. And it may finally be getting there.
But there's still very little that's as robust as a text file and a filesystem. I think a happy medium would be something that enforced some kind of syntax on all config files, and provided a single interface to them when you wanted it, but still allowed them to work basically as text files.
IIRC there was an article not too long ago that explained that the main difference between Office and OO is that Office makes extensive use of lazy loading, while OO essentially hammers through loading every library it may need, which not only thrashes the disk once on initial load, but again as you (likely) swap out memory pages. My recollection was that this lack of lazy loading had something to do with cross-platform compiling and linking issues, as well as MS having extensive resources to put into optimizing Office loading that OO did not. My understanding is that Sun hasn't exactly dumped developer time into OO, either, and I believe the focus of this release was compatibility and.
Typically people solve this problem by preloading a bunch of the relevant libraries at startup, a strategy both MS and OO attempt to employ (viz OfficeStartup and OO QuickStarter). I used to detest that, but if I had 1 or 2GB or RAM and wanted to rely on OO, I might not find it so bad. I think an interesting addition to this comparison would be to see how OO fared with QuickStarter enabled, and what drain that placed on the rest of the system. Likewise disabling the JVM loading.
I also agree with you that Microsoft has built some great software in its day. Excel continues to set the standard for spreadsheets, and underpins a huge amount of business analytics. When IE5 was released it broke new ground for browser usability. They've also built some shite, and the fact that they're so insistent on the "my way or the highway" approach makes working with their software totally aggravating.
MS really does have a lot of catch-up to do in the hosting environment. My experiences dealing with Windows as a hosting server have been miserable - it's almost impossible to tell what the current configuration of some component unless it's exposed through some control panel type page, and even if you do figure out how it's configured you generally have to have an Administrator make some GUI change. That's not to suggested that *nix hosted doesn't have its own frustrations, but you can usually figure out more less what's going on. It looks like Monad represents a new direction in that sense, but they really need to do something to expose the registry via command line to make it more transparent, or better yet, ditch the registry. I'll bet the registry accounts for more than half of the unrecoverable Windows corruptions I've seen, and it's a performance bottleneck to boot. We'll see if they can pull off that kind of change.