The 10 commandments

Earlier this year Computer Weekly asked readers to provide responses to 10 questions Linux must answer if it is to become a...

Earlier this year Computer Weekly asked readers to provide responses to 10 questions Linux must answer if it is to become a serious option for corporate IT

Few companies are using Linux for mission-critical systems. When will this happen and why should I use it until it is proven in this area?
A V Le Blanc, systems programmer, University of Manchester: I think more mission-critical systems use Linux than any other. Things like firewalls, DNS servers, mail routers, etc are very unlikely to be using Microsoft products, so by this rule you ought to steer clear of them. Nearly all of our critical services run either on a Linux box, a FreeBSD box, a Solaris box or a Cisco or 3Com box. If Nasa uses Linux on machines in its satellites, what is mission-critical?

Mike Brodbelt, director of information systems: First, I would disagree with the contention that few companies use Linux for mission-critical systems. According to Netcraft, approximately 62% of Web sites run on the Apache server software and Linux is the most popular platform on which Apache is run. Most companies, particularly in the e-commerce sector, rightly regard their Web sites as mission-critical.

How can I be sure what the community is planning for the next version?
Le Blanc: The kernel development takes place fairly visibly. There are mailing lists, news groups and online pre-releases of new kernels. Most Linux-related software projects (KDE, Gnome, Reiser File System, etc) have Web sites and mailing lists as well.

When will there be proper public key infrastructure support in Linux?
Le Blanc: I am not sure what you mean by this one. Products such as PGP and SSH are about as proper as you get. There is also support for various kinds of keys in Apache, SSL and other products.

Brodbelt: PKI is usually application-dependent. In the simple cases, applications such as PGP and GPG are already in common use on Linux. The FreeS/Wan project ( is an example of an OS level IPSec implementation for Linux. The OpenSSL project ( provides software to generate x509 certificates and is an SSL enabler for many applications.

Will the Linux Standards Base really be able to hold Linux together?
Le Blanc: We have done pretty well so far. As long as the standard doesn't try to invent new solutions, we can usually reach a consensus. Despite the rough spots so far in the Linux Filesystem Hierarchy Standard, nearly all distributors try to implement it.

Cridland: If you are considering Linux in the workplace, also consider putting some time and effort into it - it will pay itself back. The LSB, and in particular with FSSTD, have already managed a huge step forward in distribution interoperability and I am sure it will continue.

Which of the various Linux suppliers will be around in two years' time?
Brodbelt: Debian will be here for certain, as it is a volunteer-run organisation that is largely unaffected by funding considerations. I would bet that RedHat and SuSE will both still be around, and probably also Caldera.

You also need to consider that your supplier joining the ranks of the dear departed is not nearly as large a problem if you have the source code for your system. In the proprietary world, supplier death is often the ultimate lock-in, but in the open source world it is far less important.

There are many companies that will support your Linux systems and would continue to do so even if the supplier of those systems was no longer around.

When will suppliers be able to guarantee 99.99% uptime officially?
Le Blanc: Never, if you mean that seriously. But, at least for some purposes, you can be close to that now. By using such utilities as the Heartbeat software, Linux Virtual Server clustering and so on, you can probably create a more or less 100% uptime system now.

Dave Cridland, Felspar: Most Linux users, myself included, consider Linux to have a higher uptime, primarily because it requires far fewer scheduled or unscheduled reboots and very rarely crashes without severe hardware problems.

How well can Linux integrate with Windows or legacy systems?
Le Blanc: Here the tricky word is "integrate". You can use Linux to serve NT systems using Samba, or Unix systems using NFS, though the Linux NFS server is by far the weakest of Linux's network services. We mount Novell filestore from Linux boxes here; all of our 10,000 users have Novell filestore.

Cridland: Linux can - if you have the patience - replace Windows fileservers and it can certainly use them relatively easily, although those carefully hidden shares are perfectly visible to it, which may alarm naive administrators. For any other legacy systems, you are probably talking Unix - at which point integration is nearly total - or else IBM Big Iron - at which point slightly less so, but still close.

What support for the new generation of Web services will Linux offer?
Cridland: The answer is, essentially, either "total, probably designed on it", or "poor, because the service in question is a closed, proprietory system". Poor often means that it has been duplicated to some extent on Linux, though. However, there is the usual caveat that you can always contribute to areas you feel need developing.

Rachel Greenham, via e-mail: Anything Java-related is fine. Sun's Java for Linux is at parity with that for Solaris and Windows, and IBM has its own version, plus Blackdown ports the JDK for non-i386 platforms. This means that anything J2EE-related and all the other pure-Java APIs (XML, JNDI, Jini, etc) are going to work out of the box. I know they do - it is my job and my development machine runs SuSE Linux.

There are no references to Linux in the Transaction Processing Council benchmarks. When can I expect some objective performance benchmarks for the operating system?
Le Blanc: I have been involved in procurements for supercomputers and other university purchases for years. "Objective" is a tricky word. There are some fairly good tests and measurements about for networking. Transaction processing has lagged behind because no-one in the Linux community was interested in it until fairly recently.

Cridland: When someone pays for them, I am afraid. If you are a large corporation with lots of cash and you want the benchmarks, then pay for them. That single action will generate a huge amount of publicity, and is something that the developers, by and large, cannot afford.

Linux has been an overnight phenomenon. How can I be sure that this will not fizzle out and leave me with a system that no-one supports?
Le Blanc: Linux is now 10 years old, though the GNU parts of it are mostly older, as are those coming from BSD and the MIT X Windowing project. Red Hat and SuSE have both been around since about 1993 and the free Debian project, with about 1,000 volunteers, is about the same age.

If you look at the money IBM, Hewlett-Packard, Silicon Graphics and others are putting into Linux, I do not think it is likely to disappear in the next few years.

Tom Potts, systems analyst: Windows 98 will no longer be supported soon and that is only three years old, not 10.

Read more on Business applications