Paul Ferris writes a great rebuttal to yesterday’s nasty ZDNET article. Update: Paul Gallegos has worked with Mr. Ferris to improve the quality of this rebuttal. It is now ready to be submitted to ZDNET.
The dissection of “weak value propositions” – a response to Ed Muth.
By Paul Ferris and Paul Gallegos
This article is an answer to Ed Muth, group product manager for Microsoft Corporation, on his recent interview regarding Linux in ZDNet’s PC Week Online forum.
Before beginning, it must be noted that Microsoft clearly admits that FUD – Fear, Uncertainty, Doubt tactics, typical Microsoft answers to superior technology – isn’t going to work against Linux (cf. “The Halloween Documents”). However, Mr. Muth seems to be attempting these tactics anyway, perhaps as a way of keeping a negative spotlight on Linux.
In the spirit of playing along, each point of the criticism is addressed below.
1) Linux lacks a broad base of support for applications.
It is true that many of the commecial applications on the market today are Windows-centric. It is also true that many name-brand vendors are just barely jumping on the Linux bandwagon. However, it is definitely not true that Linux lacks a broad base of applications support. One can easily hit sites such as freshmeat.net and browse the daily packages contained therein, Star Division provides an office suite that rivals MS Office, KDE and GNOME easily turn the command-line interface into a comfortable GUI suitable for use by novice users, and that’s just the tip of the iceberg.
Another truism: Windows 2000 (formerly NT) does not run many of the available commerical off-the-shelf (COTS) applications, but the WINE project (WINdows Emulator/Wine Is Not an Emulator) has proven with 95% efficiency now that their software will run almost any COTS application without any need for recompiling or rewriting. One could even call this porting of COTS applications making Linux backwards-compatible.
However, many businesses do not use COTS applications; instead they use home-brew software packages created by their own IT staff. It is inherently easier to write a progam when you have complete and utter access to the operating system’s source code than it is to write a program inside an OS’s interface to let you know whether or not a subroutine is allowed or not. (cf. Visual Basic, Visual C++, etc.)
An aside: the statement contradicts Paul Maritz’s testimony under oath, where he specifically laid out many of the applications available for Linux and claimed that Linux is a superb operating system that will have no problem gaining more COTS applications in the future.
2) Linux lacks a low level of integration between the OS and its applications.
Linux lacks a low level of integration between itself and Windows applications, yes. However, in terms of Linux applications, all one needs to do is look at uptime and stability – the stability of the platform is due to the high level of integration between applications and operating system, and due to the INability of an application to crash the system itself.
No, at this point one cannot embed a Star Office document into Wordperfect or Lotus on the desktop. But that’s the desktop, and as mentioned above, more and more COTS applications are being ported to Linux or already run under WINE.
In the client-server model, CORBA, which has been around since 1991, is fast becoming the defacto standard for all hardware and software platforms. Native Linux applications can take advantage of it, as do many of the newer COTS applications. Once this happens, integration between applications and operating systems will not be as important.
As an example, recently it was found that in the Debian and Red Hat versions of Linux, the lsof 4.4 package had a security hole. Two days later it was patched. In fact, the actual time between announcements was 29 hours and 15 minutes. (cf. LinuxToday.com/security)
On the flip side, security holes made known in Windows NT 4.0 will not have a patch released for several weeks – a “workaround” from CERT or another organization is used in the meantime.
Which of the two examples is more secure?
ANY system is secure if it only is a console and not networked – except for Windows 95/98 and the infamous “Cancel” option upon login. However, how many companies do not use a network, even if it is just a small 8 port LAN?
Ed Curry, the Government security expert responsible for Windows NTs C2 security rating has been sounding the alarm over the fact that Windows NT 4.0 was not C2 secure when the manual misleadingly suggested that it was. If Microsoft were concerned about security at all, they would not charge less for Windows 9x, and more for Windows NT. At around $250 for home computing that’s a bit much when a PC costs under $1000 these days. Especially when Linux’s cost is free.
Postscript is the standard for file printing. Linux has supported this for many years now. Due to Linux’s stability and scalability, a Linux print server is much more reliable than an NT print server.
3) Linux is more costly than Windows NT.
The bottom line is that many major companies look only at the total cost when purchasing servers and desktops. Linux is free, plain and simple, therefore it can never cost MORE than Windows NT/2000.
Unix in general is more reliable and scalable than NT. With less downtime, IT staffs have more time to devote to other critical projects. That means less man hours rebooting, cleaning out print queues, unsticking mail queues, file serving, disk mounting, the whole nine yards. That also means lower total cost of ownership – the IT staff spends less man hours fixing problems and putting out fires and more man hours working with the operating system and its applications instead.
“Let’s say, for discussion, they are equally scalable.” This cannot be done. Linux can be ported onto almost any hardware, big or small. Windows only runs on DEC Alpha, Intel, or Intel clone hardware. One cannot even imagine how Windows would become “equally scalable” to Linux.
However, for argument’s sake, assuming that both are equally scalable, the stability issue still exists for Windows. That alone will cause IT managers to think twice before trusting mission-critical tasks to the machine.
“And let’s assume applications are available for both and setup time is the same.” Discarding the setup time alone, which in itself is a major effort – having to reboot after every single change is not conducive to expidient work, Windows NT/2000 has a proprietary tool to manage its network’s applications, while Linux can run just about any tool (outside of NT’s) necessary to complete its task. Licensing is not an issue since the GNU Public License allows free distribution for Linux applications, unlike Windows NT/2000. Upgrades are also free, as are patches, unlike Windows NT/2000.
Linus’s recent keynote speech at LinuxWorld ’99 made an important point here. If you have a server running the older Linux kernel, and you don’t feel like upgrading, nothing is going to force you to do it.
4) Linux has no long-term development road map, and there’s a higher technical risk in using it.
There’s no long-term development road map.
Linux holds the power to harness the global potential of the Internet, presenting far more possibilities than Windows NT/2000 can ever hope for. Linux does not need a “long-term road map”. It has scores of people around the world that can make it turn into whatever they need to accomplish a given task. Contrast that with NT/2000 today. NT/2000 is inflexible, due to its outdated, closed source, development model.
There’s a higher technical risk using it.
Define the risk. If the technical risk is relegated to security, patches are created in a matter of days. If the technical risk is related to training, the learning curve is exceptional once a user masters the idea of a command-line interface instead of point-and-click – and even then, desktop managers can be used to combine point-and-click with command-line inteface.
5) Linux is not that much cheaper than NT when comparing cost per transaction.
With Linux one has so much in the way of choices for databases. Most of the large databases are either available, such as Oracle or MySQL, or coming available soon. Only one database – Microsoft Access – has not announced a Linux port of its product. Again, scalability is not an issue, since the major database programs run on the big number-crunching machines out there anyway.
Aside from databases, Linux uses TCP/IP standard to transmit and receive its data across a network. Cost for this transaction is just as small as one would get using a Solaris machine running Oracle 8. File serving uses NFS or SAMBA. Again, the cost is the same as a generic Unix machine. Finally, web hosting or mail servers use Apache and/or sendmail, with minimal to zero cost per transaction.
6) Acknowledging the movement.
“We’re in the business of wanting the customer to have the information needed to make informed choices. … Some criticalness is needed.”
Do anti-competitive, monopolistic practices count as helping the customer make an “informed choice”?
Yes, Linux kernel version 2.2 was supposed to be released “around Christmas.” Two months later, it was there. When was Cairo supposed to be released? Or NT 5.0 (now Windows 2000)?
How about bugnet.com’s lack of bug-free product award for 1998? Quoting from their web page:
BugNet’s data indicates that bug fix rates have declined with every new mass market
version of Windows. The bug fix rate for Windows 3.x (OS and apps) was/is higher
than for Windows 95, and Windows 95’s bug/fix rate was/is higher than Windows 98.
In other words, in a broad sense across the industry, a lower percentage of bugs are
being fixed with each new generation of Windows.
And this is without mentioning the FrontPage 98 bug that allows users to delete an entire hard drive without warning. “Criticalness” indeed.
“We don’t see people question the Linux numbers.”
The reason people do not question IDC’s numbers is because IDC is an independent entity that gains nothing by falsifying its statistics.
“[Microsoft feels] that 2 to 20 percent of Linux shipments turn out to be ‘shelfware’…” only says that Microsoft itself does not know the exact number of people using Linux or its applications; Microsoft is only making a semi-educated guess. However, a recent breakdown of Internet servers in Europe showed that Linux has a higher installed base there than Microsoft. An educated guess would be if IDC could get a real track on the situation, the numbers would be much higher for Linux than it would be for Microsoft.
“You have to separate out what OS they will install if you ask them and what investments they make.”
This is false. One must consider what OS the customers will ask for in addition to the investments made.
“I find it hard to believe that some of the best computer scientists in the world will want to do their work for free … I do not believe in that vision of the future.”
That type of short-sighted thinking is based upon the premise of making money, period. A glance at any research university in the country will net you hundreds of millions of graduate students doing research, creating new programs, and working simply for a degree – and while it’s not “free,” many are not getting paid for their efforts either.
By far, though, the best example is the Linux OS itself – started in 1991, it’s still around eight years later. It does not take much to realize that the phenomenon is happening despite the negative campaigning coming from Redmond.