Can Linux break through to the commercial office desktop?

LINUX BREAKING out of its commercial server place in the sun and becoming a viable desktop platform has had its proponents for quite a few years now, and some suggest this year will be breakthrough time.

Certainly there is little to stop Linux on the desktop for home use. Its use to date has been largely confined to technical users because converting from Windows to Linux takes a certain level of IT knowledge, but that is a situation that is slowly changing.

The greatest potential market is desktops in the commercial world but the reality is that there are a number of business and technical reasons that may delay this uptake and this report looks at some of the obstacles.

We won't look at the technical obstacles, more at the factors involved in any decision to move from Windows to Linux and at the obstacles to such a transition. It is written with a strong emphasis on the business perspective because this is where such decisions are ultimately made.

Let's look first at some of the factors where Linux supposedly has the edge over Windows and see if this holds for commercial desktops.

Stability
The stability of more recent versions of Windows is quite good, particularly so if the commercial business is using a standard group of commercial applications.

Many might say that Windows is unstable but my personal experience at a very large organisation is that I have had no Windows NT crashes in the last 3 years of use. According to many reports, Windows XP is even more reliable.

In our case, I attribute the system stability to having very good system managers and rigorous test procedures - which is why we bypassed Windows 2000 and are still running tests with Windows XP. When the software is to be rolled out to hundreds of users, it is absolutely vital that the more than fifty software applications run without any problems.

I stress that this is a commercial environment and not a home or small business environment.

I have no doubt that Windows systems are crashed by programs that try to operate outside the normal bounds or that contain coding errors, but in the commercial world these should not get past the testing process - and users should be banned from installing their own software.

In general, there does not appear to be a huge difference in stability of the two systems when it comes to commercial work and stability alone is not a terribly persuasive reason for Linux to displace Windows.

If there is a good case for replacing Windows with Linux purely on the grounds of stability, then it becomes a business decision as to whether the current Windows environment should be corrected - if it can be - or whether the move to Linx should proceed despite the numerous costs and other difficulties that this might involve.

There is no "right" answer for all situations, no "one-size-fits-all" and each commercial organisation will make decisions based on their own evaluation of costs and benefits.

Security
The security of computer systems is absolutely vital to most commercial enterprises and while there are general steps that can be taken, the security of the operating system itself is an important part of the whole.

The Bugtraq database reports that in the last 15 months there have been a similar number of vulnerabilities reported for Linux and Windows XP Pro. Reports just last week of vulnerabilities in both products emphasised this.

Certainly there have been many such bugs reported for Internet Explorer, but like bugs in all applications on all platforms, if you don't run that application then you don't have a problem with that bug.

The number of Bugtraq reports can't be taken at face value because there is the potential for Linux bugs to have multiple entries in Bugtraq and the severity of any bug in a particular IT site may be quite different to the severity for another site.

No commercial business should be operating without a firewall to prevent or at least minimise the potential for damage, and this greatly reduces vulnerabilities.

Aligned to the notion of security is the need for software patches to correct the vulnerabilities that are reported. Patches are required for Linux, just as they are required for Windows.

Linux aficianados claim that the Linux world responds quickly when vulnerabilities are reported and that's certainly correct. This is because the testing of application software after these Linux patches are applied is left to users of the OS.

Microsoft has a reputation for being rather slow to provide patches for certain software problems. On the other hand these patches are probably tested against a whole range of Microsoft applications before the patch is released to users.

Which approach is better? It depends on the risk that a company is prepared to take and how much effort it wants to put into software testing. If a fast solution is needed and the risk of application problems is acceptable, then the Linux method is suitable. Some companies are quite satisfied to wait for the fully tested patch especially if their business is not being badly impacted by the problems or they lack the skilled staff to manage them.

No discussion of security would be complete without some comment about Windows XP and how it attempts to send data back to Microsoft. No commercial organisation would tolerate this data transfer when competitive information may be at stake.

Confidence in the Supplier
It is very important to a commercial business that it has confidence in its IT suppliers. It is exceptionally reassuring that the computer companies are in a position where they can throw resources at an urgent problem. The last thing a company wants is to be losing money while they wait for a solution to the problems, and this applies regardless of whether the computer is some huge system in a computer room or a desktop system that runs some vital application.

At the moment this kind of support for Linux is not easy to find and holds back potential customers.

Who would one call if a problem was found on a Linux system? Would it be the application supplier, the provider of the network device, some hardware supplier or the writer of the device driver? This problem needs a solution because a company is not going to enjoy being bounced from one help-line to another. If they are told that it is a problem in Linux - what then? What is the contact point for Linux problems?

IBM - and more recently HP - have recognised this problem and are making a stronger commitment to the support of Linux so they can provide this single point of contact for problems on their Linux systems. Many large companies find their involvement to be reassuring but are waiting for proof of that commitment before jumping.

The unfortunate downside of this is that computer companies such as IBM, Dell, HP, Sun and even Oracle others may "fork" the Linux code base and modify it to suit themselves, thus creating exactly the same problems that Unix suffers from. The chances of those companies optimising some part of Linux to suit their own hardware is very real especially if it gives them a significant edge either in performance or reliability.

It really is a "no-win" situation for Linux because the business world wants strong support for all its IT systems. Certainly these computer companies will set a price for this kind of support but such a price is regarded as "IT insurance" by customers and for most, this is well worth having.

Software Costs
It might seem odd that cost is a factor when Linux is free but it would be a strange computer system that runs no application software and has no costs associated with any aspect of support.

For most IT environments the cost of acquisition is usually far less than the cost of support. These support costs include the licence fees and maintenance fees paid to external suppliers and the costs of in-house support for the IT systems.

Linux might be free, but there are costs associated with the operation of almost all systems based on its use. If a business wants the reassurance of dealing with a major IT vendor for its Linux, this always comes at a price. If a company wants this support to be in-house, then it pays a premium for the skills and it may risk its entire operation on the continued employment of people with suitable skills.

At the application level, things are not so different. Packages like Star Office were free for a while, but no longer. Other Linux packages vary in price. Some are free or almost free but other software from commercial sources will often be priced similar to their counterparts on Unix or Windows, reflecting commercial reality. No software house can act like a charitable institution and stay in business.

When the cost of Windows at a company level is annualised over several years it reduces to just a few hundred dollars per desktop - abou the cost of a skilled engineer on site for an hour or two. With good support practices in place, the effective annual cost of these desktop systems can be kept quite low. Factor in the various enterprise discounts that are available for software and it appears that the difference in cost between Windows and Linux is not so great.

There will be some potential users where the total costs on Linux are much lower than the costs for an equivalent Windows system and for them it might well be worth considering making the switch.

Obstacles to Moving
Here are some of the problems that large commercial organisations are likely to encounter making the move.

Few of these problems can be considered trivial and so time, effor and money are required.

Installation
The installation of Linux as a replacement for Windows is not a trivial exercise in a large commercial organisation.

Each PC will require individual attention and even though special installation CDs can be used, there will be some amount of manual intervention required on all systems. This might be as little as triggering the installation from the CD or as much as setting a number of individual parameters on the PC. Afer installation, there will also be a need for some rudimentary testing even if major testing has been completed at a central location.

It might also be desirable to complete the transition as quickly as possible and this would require a large number of staff with the necessary skills to make the changes.

Most large companies are spread across a number of physical locations and so the logistics are even more complicated.

By far the easiest approach to all this would be to introduce Linux at a roll-out of new PCs across the organisation. Here, the installation of Linux can take place in a centralised location, all testing be done prior to the PC reaching the user's desk and the dispersal of those new systems ensure a rapid change-over.

Continuing in-house support
After the installation, there's support to be considered, preferably from a centralised support group.

Well-organised Windows sites have management tools that can handle the roll-out of new applications or application upgrades, and they can do it in the evening or overnight. Similar tools need to be in place for Linux support and of course the skills to use those tools must be acquired.

If the old Windows support staff are being retained, then they will need to acquire the corresponding skills for Linux. There are conceptual differences between the operating systems and much will depend on the level of support that is provided and the breadth of knowledge that those people are expected to have.

The most practical approach to this appears to be to acquire a core team of highly skilled Linux people and to retrain existing support people who are already familiar with the business and how it operates.

External Support
As pointed out before, external support for Linux can be a thorny problem for any large commercial enterprise.

Major companies want the peace of mind that some other company can deal with the messy work of fixing software bugs, testing the changes, and then providing software patches that will/should work.

This kind of external support for Linux is not so easy to find. It is improving -- with the major PC computer vendors becoming more closely involved -- and this will provide the single point of contact that most companies prefer.

The problem right now is having confidence that these vendors can actually deliver the Linux support goods. Until that's proven to be as good as the quality in other IT support, it is a real obstacle to the widespread introduction of Linux on the desktop, or in fact into any vital business areas, for most companies.

Software Replacement
The removal of Windows will also mean the removal and replacement of all software that uses APIs or DLL files, incompatible with Linux.

In most cases Linux software that is compatible with the standard file formats can be obtained, but often the range of choice is limited and the execution of that software may be quite different. If alternative software is available but file formats need to be altered then some tough decisions will need to be made.

In some cases here we also strike the problem of "business crediblity" because the alternative software might only be available from a small software house and a commercial enterprise may be reluctant to risk dealing with a company that might disappear. There are no guarantees that any large software company that writes software for Windows will also continue to exist but the risk of them failing is perceived as being rather less.

Hardware replacement
Before we look at the impact on users, there are certain peripherals designed specifically to work with Windows and drivers for these are only available via Microsoft. If a site has any of these and wants to move to Linux, there is money to be paid for the replacement hardware - and the disposal of old hardware to deal with - or some alternative service needs to be provided.

User acceptance
The ultimate test of software and the operating environment is whether users are happy working with it.

Although notionally equivalent products might look and feel like Windows, users may be uncomfortable if there are too many differences that they must quickly become familiar with. Compound this with a number of different applications which may suffer from the same problems and the effort involved retraining these users can become significant.

It is not simply the cost of retraining that hurts because productivity might be seriously compromised until the new skills are acquired.

These problems will not apply if users run applications that have the same user-interface on Windows and Linux, and the operating system is entirely invisible. Some applications like this do exist but run a risk of legal action for "look and feel". Odd this. Windows interfaces have been touted for many years as being "standard".

Some Scottish police forces have moved to Linux-based systems recently, and given the nature of their work, it's to be expected that they have just a few, quite special bespoke applications. This represents the ideal case for moving from the Windows to the Linux OS, as modifications to software are entirely invisible to users and they'd need no re-training.

Summary
Linux is coming to the desktop, there is no doubt about that at all. But the speed at which large commercial enterprises move to Linux - or in fact, if they move to Linux at all - will be predicated on the business case that is involved with any such move, more so than any technical reasons.

There are some steps that the Linux community can take to improve the likelihood of big companies shifting to the OS, but some of those steps may not be entirely compatible with Linux philosophy.

It's initial and continuing costs, along with potential business risks, that will ultimately decide whether big firms jump or stay with what they've got. µ