Mr. Jacobowitz talks about standards with representatives of the various distributions by e-mail and at the LinuxWorld Expo.
With all of the recent debate about standardization and the future of Linux standards, it's natural to wonder where the major Linux distributions and other key industry players stand on the subject. So, to find out, Linux Journal posed the following question: “What stance does your organization take on the subject of Linux standards and the growing standardization issue?” We received some very interesting responses with quite a variation in length from the major vendors. Here they are, in no particular order.
We at Red Hat are generally in favor of good standards that speak to existing as well as future problems. History has shown that adherence to standards can be very important and failing to do so can cause fragmentation that leads to a very unhealthy situation for all players involved. We do our best to make sure we adhere to useful standards such as the FHS (Filesystem Hierarchy Standard).
New standards for Linux are evolving, and we are working hard to be a part of that evolution. Standards are very important in maintaining compatibility across all Linux distributions. Compatibility is important to keep independent software vendors (ISVs) from seeing Linux as a fragmented mess that can't be supported. ISVs are important because they bring the applications to Linux. Applications are important because they bring users to Linux. As you can see, it's a trickle-down effect that has large implications.
That said, we are working hard to make sure the evolving standards are good ones. Bad standards can also be very harmful. So far, the Linux community has been very good at working out good standards. With a responsible community effort from all players, I fully believe the Linux community can maintain that trend. If so, you can expect Red Hat to participate fully with the standards that are created.
Thank you for the opportunity to respond on Caldera Systems' feelings about Linux Standard Base (LSB). Everyone probably has a slightly different definition of what they think LSB is and what it should be. I will take the liberty of telling you what I think it should be. LSB's main objective should be to solve the problem facing commercial ISVs. LSB should be an effort that would enable commercial software developers to port once and have their applications run on all major Linux distributions. If it does not solve this problem, Linux will have a major problem in the not too distant future. Nearly every ISV we have spoken to wants to remain “distribution agnostic”, but they do not want to support four or more versions to accommodate Linux. The current answer seems to be one of three options: publish the source code to the application and allow the community to support the different versions, pick a kernel version and a library version and allow the Linux providers to offer patches, or just pick one of the distributions to support.
I will address the problems with all three of these ideas. First, publish the source code to the application. This will work in some cases but not in the majority. Some of these industry players have millions if not billions of dollars tied up in proprietary software and hardware that will not allow them to publish their solutions. They do and will continue to support Linux and the open-source effort, but their board of directors, attorneys and stockholders will not allow them to publish source code at this time and may never allow it. Publishing the source opens up significant liabilities. Netscape published the browser source only to outsource it to another company. Publishing source has to be done after significant thought and consideration.
As to the second option of allowing the ISV to pick a version of the kernel and library, what if Oracle picks one and Informix picks another? Or if Corel and Oracle pick different versions and the customer cannot run one of these solutions without searching the Internet to find patches? How will Linux compete against Microsoft when they've claimed to be a one-stop shop? The commercial customer loses because he must know his way around to find the patches and fixes on the Internet. Even VARs and Systems Integrators who are technical enough to do so cannot afford to spend the time.
The third option is to port to just one distribution. This seems counterintuitive to the Linux and Open Source community that believes in choice. However, there are those who would support this option to solve the problem. The major concern with porting to only one distribution is you lose out on the innovations being made by others who are targeting different markets. Some of the leading distributions have targeted the developer as the principal customer, as developers have made up the majority of those purchasing Linux until now. Features like time-to-market are of vital interest to the developer. Packages are gathered from many different sources around the Internet to ensure the quantity and timeliness of delivery, and little concern is given to ensure that source and binaries match. After all, the developer knows enough to make it work.
Commercial concerns cannot deal with more than two releases a year, so time-to-market is not a feature but a liability. Caldera Systems has been focusing on the commercial or business market from its inception and has built features like self-hosting into the distribution. Self-hosting is the discipline to match all source to the binaries. It allows the commercial entity to set all of the compilation parameters for the entire distribution, maximizing performance and support options while limiting security and liability concerns. All the packages are tested and integrated as a single whole. Distributions focused on the developer may choose to take the time to add more packages which appeal only to the developer, or release a technology before it is stable so that developers have easier access to it. The bottom line is that the Linux community and the business customer will lose options and choice if a single distribution is used as the reference platform. Some ISVs may choose one and some another. The only real answer is for the Linux community and the current providers of Linux to collaborate.
Linux and the whole Open Source community have a great opportunity to become significant players in the commercial environment if they remain open-minded. If Linux and the Open Source community can make it easy for commercial players to interact with Open Source without incurring legal liabilities or major costs, Linux will be a significant player and the customer will win.
LSB can be a major step toward solving this problem and showing the industry that Linux is different, not only because it is published under an open-source license, but because its providers have vision. If the major Linux distribution providers can collaborate and support a common kernel and library version as well as the basic interfaces that enable an ISV to port once to Linux and support all Linux providers, everyone wins. If the Linux providers are irresponsible and choose not to cooperate but continue on with their own agendas, Linux is not as likely to fragment, but will fall significantly short of its potential to be the predominant computing platform, open source or otherwise, in the industry.
Linux is not as likely to fragment, as its cousin UNIX did in the OSF/UNIX International debacle, because it is Open Source. However, being Open Source will not completely solve the problem either. The premise that the best open-source technology always wins is not true. GNOME and KDE are prime examples. KDE is excellent technology that has matured faster than GNOME. The main reason GNOME was started was because the Qt libraries KDE depends on were not Open Source. Since Troll Tech has now published the libraries under an accepted open-source license, all concerns about KDE should be over. Instead, a significant public relations effort has begun to continue to spread uncertainty and doubt on KDE. Technologically, it is superior in all cases and is more mature. Why would someone persist in promoting false information about KDE? Because they have invested time, money and energy in GNOME and the KDE developers do not have the funds to fight back. Again, the Open Source community is not immune to the influences of the almighty dollar.
Clearly, if Linux is to achieve its potential as a major industry alternative, we cannot rely upon Open Source alone. All of the industry players must act responsibly. Caldera Systems has always wanted Linux to be a viable commercial alternative, so we believe the only real answer is to support LSB.
Pacific HiTech will take an active role in the development of Linux standards and will make compliance with the adopted standards a design goal of all our Linux products. We believe a well-defined set of Linux standards is important so that ISVs can more easily port their applications to the Linux operating system without worrying about distribution compatibility.
I'd like to see them develop to the point where any compliant distribution will have the same shared libraries available and the same basic file-system structure. Some people have suggested that a standard package manager would also be important, but I'm not necessarily sure that's true—for example, .deb and .rpm can both provide a system in which a given third-party application could work, provided that the shared libraries and path structure are the same. What I would like to see is a common system for “registering” applications so that desktop managers and other programs would have a standard method of determining what is installed on the system.
Slackware is waiting to see what the proposed standards are before we commit to complete compliance, but I do think the effort is a step in the right direction for Linux. I'd really like to see a list of standard version numbers to use when building shared libraries—that should be a simple first step in getting binary compatibility among the distributions back on track. However, the standards should probably not go into such detail that all distributions end up looking cut from the same cookie cutter. It is the different design philosophies which make different Linux distributions appeal to different kinds of users. It'll be interesting to see how these issues are balanced.
The general stance of Debian is that standardization is a good thing. Dale Scheetz is a Debian developer and actively involved in the LSB work. LSB is also an SPI-supported project. As Linux continues to grow, it will become more and more important for vendors producing software to know it will run on a number of different distributions without requiring distribution-specific versions. The growth of Linux will be greatly helped by applications being ported to run on it, and this growth must be supported as much as possible.
Debian has been involved in the LSB project from the beginning. One of our developers, Dale Scheetz, is working on the LSB right now. We've been talking with Red Hat since before LSB so that we can develop binary compatibility between the distributions. The important system libraries should be fully compatible across all distributions. We don't want to see the kind of incompatibilities suffered by users of 16-bit MS Windows software when they upgraded to 32-bit MS Windows. Debian is currently working to adopt the File Hierarchy System, but we feel a few issues remain to be resolved. Debian is happy to help create standards and compatibility with the other distributions; the LSB is one of the methods we are using to make this happen.
SuSE has been a member of the Linux community since 1992, and we have people dedicated to and working on the LSB project. We are proud to be active members of this project. We are hoping to see some sort of minimum library standards because that is what our customers want. ISVs need that kind of standard to work efficiently. Hardware vendors deserve a standard in order to certify their hardware across distributions.
It's my personal opinion that standardization is good thing. But like always, too much of a good thing can be a bad thing. It really all depends on which aspect of the Linux system you want to standardize, and how much. Where do you begin and where do you stop? If we standardize the file system layout, packaging and user tools, is anything left? I'm not sure of LSB's plan—but from my reading, it doesn't look like they have a well-laid-out plan quite yet. Standardizing file locations is a great idea; I know it would solve half of the problems I encounter. I read somewhere LSB plans on having a standard packaging system. That is a tricky subject for me to even personally comment on—we have a lot of plans for our Stampede Linux Packaging system (SLP). There are some features the standard may not include that we'd like to have. Obviously, we can contribute to the project ourselves, but if the majority doesn't like our idea, then it won't go in. The great thing about Linux is that if you don't like something, you can't complain. Don't use it, find something else, hack the source code, or start a whole new project. If a situation like this arises, the user who wants a feature can no longer search for another package; he must try and somehow convince the authors to add it.
At this time, two extremes are apparent: one is “everyone does their own thing, their own way” and the other is “everyone does what the standard says”. I'd like to think we aren't at either of these extremes right now. That fine line between them needs to be found, and when someone finds it, I'll be able to comment further.
Currently, with the information we have, Stampede's stance is neutral, but this could easily change. It could change tomorrow, next week, next month or next year (just like any Linux-related project). For now, I will try to keep a close eye on what goes on and keep others of the Stampede development team informed. It will be much easier to make a decision once a clearly defined set of rules has been set.
