Both of those statements are true. Linux has never been more user friendly, and it's still a pain in the ass. The way I see it, a mainstream Linux desktop is a few years off, just like cost-effective nuclear fusion generators.
Cross-environment compatibility is generally achieved by reducing integration with the desktop environment. That results in a lot of wasted work and a less polished end result. Evolution doesn't have this particular problem, as it is very much a GNOME app. It does, however, integrate much more nicely with an all-GNOME desktop than it does with a KDE desktop.
MS doesn't have this problem because they have a single target to which one can develop. Win32 certainly does evolve over time, but it doesn't branch off into redundant, incompatible variants. The Mac Office group does have to support a completely different platform, but they tailor their app to the Mac desktop instead of trying to abstract the differences between Mac and Windows down to a lowest common denominator.
One feature of Outlook/Exchange that doesn't get mentioned very often is the integrated security and delegation system. It's quite easy for users to grant and restrict access to portions of their account, and they can do it straight from the mail client. Most users seem to be able to figure it out without any help.
I did a test deployment of Bynari Insight Server (Cyrus+exim+OpenLDAP+web interface) once, and found that the equivalent capabilities were very limited and difficult for users to understand. And that was the least of the problems we would have faced in a migration. I do hold out hope for Samsung Contact, which I'll probably be evaluating this spring. That seems to be a much more integrated solution that was designed with the business user in mind.
I can see this capability being useful for the home user and small companies, but I can't imagine that any medium or large corporation is going to want internal communications to be distributed like that. Pushing complexity out to the desktop makes it more difficult to provide IT support to your users. Workstation setup, networking, change control and backups all become much harder when every machine is a server.
Anyway, it remains to be seen how Chandler is actually going to do the P2P thing, but as an IT manager I would definitely want the option to centralize data storage and security configuration.
You remind me of the Windows developers I had to support a few years ago when I ran a helpdesk. Sure, storage is cheap...if you don't need to worry about IT support, redundancy, reliability, backups, or high-volume network access. If you do need any of those things (and I imagine that the Galaxies servers will need all of them), then the price goes way up.
If it were really that easy to reverse-engineer a binary driver as complex as Nvidia's, then their competitors would have better drivers. As it is, driver quality is a big point in favor of buying Nvidia cards. ATI, in particular, could learn a thing or two from them.
In the experience of most other people, Office is at least better than any of the alternatives. I've managed to switch entirely over to Openoffice at work, but I mostly stick to basic features of the word processor and spreadsheet. If I ever have to do anything complicated (particularly if the resulting file has to be shared with other people), I'd have to go with Office. Even in WINE it runs rings around Openoffice, and is just a more polished product.
...assuming that you have all of the required headers and libs in the correct versions and locations. And that the source code is happy with your versions of GCC, Make and the operating system.
"./configure;make;make install" works without additional effort for me about 70% of the time on a system where I keep development packages reasonably well updated. The rest of the time I can usually sort things out by hand, but I'd really rather not waste the effort if it's for personal use.
There's also the minor problem that compiling any reasonably sized application is *slow*. On production systems I'll usually build an app from source once, then package it up for binary distribution. Source installs are for development environments.
The difference is that open source projects are much more dependent on community input to uncover bugs. Commercial vendors have QA departments who bug hunt for a living. They're generally a lot more efficient about it than a typical user, and have a closer relationship with the developers.
Of course, when it comes to uncovering obscure bugs across a wide range of environments, mass user input is very useful. But you really want to get the obvious kinks out of the product before you ask Joe User to take it for a spin. Otherwise you're going to get countless duplicate reports for bugs that the developers already know about.
In the Linux world, we count on the distro vendors to provide that layer of professional quality control before and after a release. Unfortunately, their resources are rather limited at this point.
Except that the author is free to reissue his code under a different license. I would say that the GPL restricts the "users who are developers"'s rights, in order to maintain the "users of the users who are developers, who are also developers"'s rights.
I think you're using the Fox News definition of "evil", as in "something I disagree with." As a former OS/2 user, I've got a long history of bitching about Microsoft, but at its worst I would only call the company ruthless and somewhat unethical. I would save the term "evil" for organizations that engage in truly immoral acts, like (for instance) knowingly getting people killed.
Outside of Microsoft and the software industry, I think Gates has led a fairly admirable existence. Of course it's easy to be generous when you're richer than God, but he's still giving away money willingly. More importantly, he's intelligently using his wealth to try to make a lasting positive impact on human society. What more could you ask of the man?
You are, of course, welcome to contact the Gates Foundation's beneficiaries and inform them that Bill's other baby makes products that don't always interoperate with its competitors' products. I'm sure they'll start sending the blood money back.
There's a tradeoff here. It's definitely bad to enforce backward compatibility since the dawn of time. But with the Linux driver model you can't even assume compatibility between minor revisions of the same kernel. That's disruptive for both users and driver developers.
I kind of like Apple's approach to backward compatibility. Every few years they make some major change that pushes everyone to replace a bunch of hardware and software (e.g., the PowerPC switch or Mac OS X). But in between you can count on a reasonably consistent environment. And you can usually still use your old stuff with some minor inconvenience.
Your post is basically on target, but I do have one little nitpick: I would start by laying off the guy who replaced Exchange with a Sendmail process, because he clearly doesn't understand that Sendmail is only an MTA and not a drop-in replacement for Exchange.
It's been a while since I've studied this, but isn't that primarily a mainstream Protestant view? There are certainly many Christians who believe that good deeds are in fact relevant to one's eternal fate. That's why I buy indulgences directly from the Vatican.
Anyway, I think the salvation through faith thing is even worse than an arbitrary code of morals handed down by God. Why not just send all the Pepsi-drinkers to hell and reserve heaven for Coke-drinkers?
ACLs are most useful to users who are not administrators. For instance, they allow an end-user to grant additional permissions to another department, or a user in another department. This can be accomplished with ugo permissions, but it requires users to submit a request to the IT department (thereby ensuring that it will never get done), and it clutters the group file with lots of random groups (especially bad because you can't nest groups).
This argument, of course, only applies to home directories and common fileshares. I do agree that traditional Unix permissions are much more maintainable on filesystems that house the operating system and applications. As usual, what's good for the sysadmin is often not what's good for the user.
One thing Windows is much better at is supporting drivers that were released after the operating system and therefore could not be bundled. System detects new device, user inserts CD, and a few minutes later everything works. I boot back to '98 occassionally to play games, and despite four years of hardware upgrades I haven't had any need to do a full OS upgrade.
Ideally, Linus ought to lock down the kernel module binary interfaces for the duration of a stable kernel's lifecycle. That would make it easy to ship binary driver CDs along with new hardware. As an alternative (both options could coexist), someone could define a universal method for packaging driver sources (something like SRPM, but distro-agnostic) and have the OS automatically build and install new modules.
I like Frasier, and it does have a better recipe for mass appeal, but most of the time the writing isn't nearly as good as Buffy's. I don't watch soap operas or teen shows, so I can't make any comparisons with them.
It's more like 40%. And nearly everyone who uses Exchange is making heavy use of shared calendaring. Anyone who has tried to pitch an open-source alternative (myself included) has learned that "we don't support that feature, so it must not be important" isn't the most convincing argument.
1. I think people do tend to notice the difference between native Outlook/Exchange MAPI communication and IMAP. With MAPI, messages just show up immediately. With IMAP, there's still a sense of having to synchronize with the server. The users might be hard-pressed to tell you exactly what's wrong, but they do sense that things aren't working quite as smoothly.
3. At a previous workplace, we had Sendmail piping every relayed message through a virus filter program. Easy enough to set up, but it was horribly inefficient. I think Sendmail, Inc. has implemented a real content filtering API, but I haven't had a chance to play with their commercial products.
You make it sound as if the dearth of decent drivers were a small matter. It's the second biggest obstacle to Linux's acceptance (lack of certain types of software is probably the biggest), and poor support for binary drivers is compounding the problem.
I doubt it would make much difference. id's business is making 3D engines. It has been a while since their games have been compelling enough that anyone would change operating systems just to play them.
Anyway, Carmack has already been burned by the Linux retail game market. You'd have to give id a lot of money to make a Mac-only game. And even more money to make a Linux-only game. And you'd probably have to pay all of the production, distribution, marketing and technical support costs.
In order to become a mainstream gaming platform, Linux will probably have to go through a number of changes that simply won't fly with the hardcore users who do most of the development. Off the top of my head:
1) One distribution (probably Red Hat) in use by 90% or more of home users.
2) Windows-style detection and installation of binary drivers. And a reasonably stable binary driver API.
3) One desktop environment reigning supreme. Doesn't matter which.
4) OpenGL 2.0, and much faster promotion of standards by the ARB. Or native DirectX support at the same level as Windows.
5) Widespread hardware support for OpenAL.
6) Better support for USB controllers, including force-feedback devices.
7) Other stuff I'm not thinking of because I need to go to sleep.
Anyway, I do most of my work on Linux, but if wanted to develop commercial games for a Unix platform I would have to choose Mac OS X.
I worked at a Kleiner-Perkins-funded startup, and had friends working at others. In each case, you had bright founders who understood the company's core competency pretty well, but had no idea of how to run a business with more than twenty people. Nobody was too concerned with this, because the venture capitalists were supposed to bring in the resources and business expertise to turn a good idea into a huge success.
As we all found out eventually, things weren't quite that simple. Typically the VCs might drop $30 million on a company and tell the CEO that he had to hire 200 people in nine months. Or (in the worst case I know of), they would bring in a new CEO who hired a circle of cronies and attempted to squeeze as much money out of the company as possible before it hit ground.
Not that the starry-eyed founders were innocent, of course. For the most part, they were all too willing to close their eyes and abdicate real responsibility. Early on in the process, they had their chances to step in and right the ship. But they were too busy looking for the big IPO payoff that everyone was promised.
Maybe things were different with other VC funds, but KPCB was the firm that everyone trotted out as the big example of dot-com success. I even had an investment banker friend tell me that my company would be fine because "Kleiner will take care of you."
Slashdot Mirror
Both of those statements are true. Linux has never been more user friendly, and it's still a pain in the ass. The way I see it, a mainstream Linux desktop is a few years off, just like cost-effective nuclear fusion generators.
Cross-environment compatibility is generally achieved by reducing integration with the desktop environment. That results in a lot of wasted work and a less polished end result. Evolution doesn't have this particular problem, as it is very much a GNOME app. It does, however, integrate much more nicely with an all-GNOME desktop than it does with a KDE desktop.
MS doesn't have this problem because they have a single target to which one can develop. Win32 certainly does evolve over time, but it doesn't branch off into redundant, incompatible variants. The Mac Office group does have to support a completely different platform, but they tailor their app to the Mac desktop instead of trying to abstract the differences between Mac and Windows down to a lowest common denominator.
One feature of Outlook/Exchange that doesn't get mentioned very often is the integrated security and delegation system. It's quite easy for users to grant and restrict access to portions of their account, and they can do it straight from the mail client. Most users seem to be able to figure it out without any help.
I did a test deployment of Bynari Insight Server (Cyrus+exim+OpenLDAP+web interface) once, and found that the equivalent capabilities were very limited and difficult for users to understand. And that was the least of the problems we would have faced in a migration. I do hold out hope for Samsung Contact, which I'll probably be evaluating this spring. That seems to be a much more integrated solution that was designed with the business user in mind.
I can see this capability being useful for the home user and small companies, but I can't imagine that any medium or large corporation is going to want internal communications to be distributed like that. Pushing complexity out to the desktop makes it more difficult to provide IT support to your users. Workstation setup, networking, change control and backups all become much harder when every machine is a server.
Anyway, it remains to be seen how Chandler is actually going to do the P2P thing, but as an IT manager I would definitely want the option to centralize data storage and security configuration.
You remind me of the Windows developers I had to support a few years ago when I ran a helpdesk. Sure, storage is cheap...if you don't need to worry about IT support, redundancy, reliability, backups, or high-volume network access. If you do need any of those things (and I imagine that the Galaxies servers will need all of them), then the price goes way up.
If it were really that easy to reverse-engineer a binary driver as complex as Nvidia's, then their competitors would have better drivers. As it is, driver quality is a big point in favor of buying Nvidia cards. ATI, in particular, could learn a thing or two from them.
In the experience of most other people, Office is at least better than any of the alternatives. I've managed to switch entirely over to Openoffice at work, but I mostly stick to basic features of the word processor and spreadsheet. If I ever have to do anything complicated (particularly if the resulting file has to be shared with other people), I'd have to go with Office. Even in WINE it runs rings around Openoffice, and is just a more polished product.
...assuming that you have all of the required headers and libs in the correct versions and locations. And that the source code is happy with your versions of GCC, Make and the operating system.
"./configure;make;make install" works without additional effort for me about 70% of the time on a system where I keep development packages reasonably well updated. The rest of the time I can usually sort things out by hand, but I'd really rather not waste the effort if it's for personal use.
There's also the minor problem that compiling any reasonably sized application is *slow*. On production systems I'll usually build an app from source once, then package it up for binary distribution. Source installs are for development environments.
That's only if you're using them with tag switching. VLANs are still useful even if they don't share cable.
The difference is that open source projects are much more dependent on community input to uncover bugs. Commercial vendors have QA departments who bug hunt for a living. They're generally a lot more efficient about it than a typical user, and have a closer relationship with the developers.
Of course, when it comes to uncovering obscure bugs across a wide range of environments, mass user input is very useful. But you really want to get the obvious kinks out of the product before you ask Joe User to take it for a spin. Otherwise you're going to get countless duplicate reports for bugs that the developers already know about.
In the Linux world, we count on the distro vendors to provide that layer of professional quality control before and after a release. Unfortunately, their resources are rather limited at this point.
Except that the author is free to reissue his code under a different license. I would say that the GPL restricts the "users who are developers"'s rights, in order to maintain the "users of the users who are developers, who are also developers"'s rights.
I think you're using the Fox News definition of "evil", as in "something I disagree with." As a former OS/2 user, I've got a long history of bitching about Microsoft, but at its worst I would only call the company ruthless and somewhat unethical. I would save the term "evil" for organizations that engage in truly immoral acts, like (for instance) knowingly getting people killed.
Outside of Microsoft and the software industry, I think Gates has led a fairly admirable existence. Of course it's easy to be generous when you're richer than God, but he's still giving away money willingly. More importantly, he's intelligently using his wealth to try to make a lasting positive impact on human society. What more could you ask of the man?
You are, of course, welcome to contact the Gates Foundation's beneficiaries and inform them that Bill's other baby makes products that don't always interoperate with its competitors' products. I'm sure they'll start sending the blood money back.
There's a tradeoff here. It's definitely bad to enforce backward compatibility since the dawn of time. But with the Linux driver model you can't even assume compatibility between minor revisions of the same kernel. That's disruptive for both users and driver developers.
I kind of like Apple's approach to backward compatibility. Every few years they make some major change that pushes everyone to replace a bunch of hardware and software (e.g., the PowerPC switch or Mac OS X). But in between you can count on a reasonably consistent environment. And you can usually still use your old stuff with some minor inconvenience.
I think it's the "we die" part that gets people worked up.
Your post is basically on target, but I do have one little nitpick: I would start by laying off the guy who replaced Exchange with a Sendmail process, because he clearly doesn't understand that Sendmail is only an MTA and not a drop-in replacement for Exchange.
It's been a while since I've studied this, but isn't that primarily a mainstream Protestant view? There are certainly many Christians who believe that good deeds are in fact relevant to one's eternal fate. That's why I buy indulgences directly from the Vatican.
Anyway, I think the salvation through faith thing is even worse than an arbitrary code of morals handed down by God. Why not just send all the Pepsi-drinkers to hell and reserve heaven for Coke-drinkers?
ACLs are most useful to users who are not administrators. For instance, they allow an end-user to grant additional permissions to another department, or a user in another department. This can be accomplished with ugo permissions, but it requires users to submit a request to the IT department (thereby ensuring that it will never get done), and it clutters the group file with lots of random groups (especially bad because you can't nest groups).
This argument, of course, only applies to home directories and common fileshares. I do agree that traditional Unix permissions are much more maintainable on filesystems that house the operating system and applications. As usual, what's good for the sysadmin is often not what's good for the user.
One thing Windows is much better at is supporting drivers that were released after the operating system and therefore could not be bundled. System detects new device, user inserts CD, and a few minutes later everything works. I boot back to '98 occassionally to play games, and despite four years of hardware upgrades I haven't had any need to do a full OS upgrade.
Ideally, Linus ought to lock down the kernel module binary interfaces for the duration of a stable kernel's lifecycle. That would make it easy to ship binary driver CDs along with new hardware. As an alternative (both options could coexist), someone could define a universal method for packaging driver sources (something like SRPM, but distro-agnostic) and have the OS automatically build and install new modules.
I like Frasier, and it does have a better recipe for mass appeal, but most of the time the writing isn't nearly as good as Buffy's. I don't watch soap operas or teen shows, so I can't make any comparisons with them.
You really shouldn't be eating Doritos in the server room.
It's more like 40%. And nearly everyone who uses Exchange is making heavy use of shared calendaring. Anyone who has tried to pitch an open-source alternative (myself included) has learned that "we don't support that feature, so it must not be important" isn't the most convincing argument.
1. I think people do tend to notice the difference between native Outlook/Exchange MAPI communication and IMAP. With MAPI, messages just show up immediately. With IMAP, there's still a sense of having to synchronize with the server. The users might be hard-pressed to tell you exactly what's wrong, but they do sense that things aren't working quite as smoothly.
3. At a previous workplace, we had Sendmail piping every relayed message through a virus filter program. Easy enough to set up, but it was horribly inefficient. I think Sendmail, Inc. has implemented a real content filtering API, but I haven't had a chance to play with their commercial products.
You make it sound as if the dearth of decent drivers were a small matter. It's the second biggest obstacle to Linux's acceptance (lack of certain types of software is probably the biggest), and poor support for binary drivers is compounding the problem.
I doubt it would make much difference. id's business is making 3D engines. It has been a while since their games have been compelling enough that anyone would change operating systems just to play them.
Anyway, Carmack has already been burned by the Linux retail game market. You'd have to give id a lot of money to make a Mac-only game. And even more money to make a Linux-only game. And you'd probably have to pay all of the production, distribution, marketing and technical support costs.
In order to become a mainstream gaming platform, Linux will probably have to go through a number of changes that simply won't fly with the hardcore users who do most of the development. Off the top of my head:
1) One distribution (probably Red Hat) in use by 90% or more of home users.
2) Windows-style detection and installation of binary drivers. And a reasonably stable binary driver API.
3) One desktop environment reigning supreme. Doesn't matter which.
4) OpenGL 2.0, and much faster promotion of standards by the ARB. Or native DirectX support at the same level as Windows.
5) Widespread hardware support for OpenAL.
6) Better support for USB controllers, including force-feedback devices.
7) Other stuff I'm not thinking of because I need to go to sleep.
Anyway, I do most of my work on Linux, but if wanted to develop commercial games for a Unix platform I would have to choose Mac OS X.
I worked at a Kleiner-Perkins-funded startup, and had friends working at others. In each case, you had bright founders who understood the company's core competency pretty well, but had no idea of how to run a business with more than twenty people. Nobody was too concerned with this, because the venture capitalists were supposed to bring in the resources and business expertise to turn a good idea into a huge success.
As we all found out eventually, things weren't quite that simple. Typically the VCs might drop $30 million on a company and tell the CEO that he had to hire 200 people in nine months. Or (in the worst case I know of), they would bring in a new CEO who hired a circle of cronies and attempted to squeeze as much money out of the company as possible before it hit ground.
Not that the starry-eyed founders were innocent, of course. For the most part, they were all too willing to close their eyes and abdicate real responsibility. Early on in the process, they had their chances to step in and right the ship. But they were too busy looking for the big IPO payoff that everyone was promised.
Maybe things were different with other VC funds, but KPCB was the firm that everyone trotted out as the big example of dot-com success. I even had an investment banker friend tell me that my company would be fine because "Kleiner will take care of you."