Commentary on "Linux in the Enterprise"

Recently I offered the essay, "Linux in the Enterprise", for publication on osOpinion.com. I had hoped for a bit of peer review, and boy did I get it. Unfortunately, I never imagined that the essay would be disseminated as widely as it was, and what followed was a deluge of e-mails criticising this-or-that point in the essay.

I was especially surprised by the professionalism of the readership on osOpinion. This essay was severely flawed, and I was braced for a flame-bath when I came home from work.. However, the flames never really came. People were supportatiove and corrected the errors in a generally pleasant manner.

What's wrong with the original essay? Lots. I won't re-print the original, since it can be found at osOpinion.com but I'll simply list here the corrections that were made in response to the many messages I've received, and a couple of comments about things I won't change. Get out your pencils, because there are a lotof corrections, and there will be a test later.

1. I've received many complaints that Linux is far too weak for the "Enterprise" for this or that reason, ranging from a single TCP/IP stack to kernel limitations. There is no way that I would deny that Linux suffers from these limitations. However, I seriously question these people's definition of "ready for the Enterprise." I certainly don't state, nor do I believe, that Linux should be used as a "one-size-fits-all" solution to every enterprise solution. Let's face it, if that were my position I would never have written that one major role of Linux is to act as a training ground for more powerful operating systems, like Solaris or AIX. My goal in writing the essay was to show that businesses should not be afraid to deploy it as a server or a workstation OS. It is simply a given that you should use the best OS for a particular job. That's why you'll routinely find mixed OS environments in most large companies. I'm currently working in an environment that includes OS/400, AIX, Novell, Windows 95, Windows 98, Windows NT, and soon-to-be... Linux. Even in my own home I don't use a single OS. I use Linux primarily, but also DR-DOS on my laptops and Windows 98 for educational programs and for developing Windows applications at home. (YES, I WRITE FOR WINDOWS!).
   
   
2. In my opinion, the biggest gaffe in this essay by far is the statement that Linux supports XFS. This was a premature statement, to put it kindly. SGI has stated that they will be supporting Linux, and that they are releasing XFS to Open Source, but the port is far from complete. This was an inexcusable goof on my part.
   
   
3. NTFS is a journaling file system. MS is just lousy at indexing their own site, so the casual surfer isn't likely to find that information without an exact match. However, if you search for the exact phrase "journaling file system" (3 words) you'll get six hits in addition to the three I noted for "journaling filesystem". I had expected to get whitepapers describing the benefits of NT. Instead these hits link to papers comparing NT to Linux and Unix.
   
   
4. The largest file size in Linux does depend on the kernel. It also depends on the hardware you run it on. We are, after all, talking about Linux, not "Lin-tel." Obviously, I should have made the distinction, and I should not have mentioned XFS here either, at least not in that manner.
   
   
5. Concerning swap files, you can't mount "as many as you like." You can mount eight. Why you would want more is a mystery. However, the 128 MB swap file size limit has disappeared. I received several e-mails from individuals running Red Hat 6.1 with swap files as large as 500MB. (I have found that on my own desktop Linux's performance improves if you give it lots of RAM and reduce the swap file size to a bare minimum.)
   
   
6. I mentioned Beowulf clustering in the section on stability. This was misleading. I should have moved this entire reference to the discussion of performance. The fact is, a Beowulf cluster running Linux kicks NT's tuchus. It was also misleading to mention NASA's requirements of stability as being superior to commercial definitions since Beowulf clusters are typically deployed in research purposes.
   
   
7. In a footnote I linked to the wrong End User License Agreement (EULA). I linked to the one for Internet Explorer 4. If you're using Windows you can view the EULA by clicking Start, pointing to Programs, and then clicking Windows Explorer. Double-click the Windows folder and then double-click the Help folder. Scroll down and double-click license.txt. You'll see that Microsoft does indeed provide a 90-day limited warranty for Windows. Now isn't that strange that they have different EULAs, since IE4 an inextricably integrated part of the OS <tic>? (Now before you get all in my face with automotive drive train warranties (and I've heard a lot of those), remember that automotive components can have separate warranties because the subsystems are clearly defined and replaceable. MS claims that IE4 can't be separated from Windows. How can there be a separate warranty for an inseparable component?)
   
   
8. Oddly, I've personally received no e-mails contesting the low TCO of Linux. I'm sure there are posts out there that attempt to refute this, but I've not had the time to wade through them.
   
   
9. I also haven't received any comments refuting my basic argument regarding security. Nevertheless, I've toned down the final paragraph a bit and removed the phrase "easy to crack." I don't spend most of my time trying to hack into NT systems that have been deliberately locked down, so I'll leave that assertion to other people.
   
   
10. I received a couple of e-mails regarding Microsoft's claim in court that Linux is a threat on the desktop. These apologists state that a being a threat is very different from being ready for Prime Time. They also point to the difficulty of configuring the system, adding new programs, etc. In this case I disagree. The enterprise situation is very different from home use for a number of reasons, not the least of which is that corporations often want to lock down the desktops. They want to discourage the wholesale installation of unauthorized programs. There's even a term for this, "Configuration Management." Corporations control the intranet, the applications, and the hardware purchased. They also tend to have qualified system administrators, which explains my "business definition of 'easy'". My mother is more fortunate than most in that she gets the benefit of that "business definition" at home. I know many people who wish they had the same for Windows.
    
    
11. I shouldn't have mentioned that I use RealPlayer G2, as it's not released yet. I'm using the alpha version I acquired from Real.com, with which I've had no problems. If this is an alpha, then I've got great hopes for the full release. However, since it's not generally available I should have left it alone.
    
    
12. I also mentioned 3D modelling and graphic design tools of a calibre used in Titanic. This was a bit of naïve hyperbole, to put it kindly. I've since been informed that there are far nicer tools out there than I'm using available on Windows. I'm not really surprised, since that's where the development effort has been concentrated. However, it has nothing at all to do with Linux in the Enterprise, and I could have left the whole thing off without damaging my argument. Curiously, I also received several e-mails taking me to task about CAD programs, even though I never mentioned any.
    
    
13. A couple of people complained that Doom and Quake were not of the same depth and quality of the newer games for Linux. Once again, this has nothing to do with Linux in the Enterprise, and my mistake was in violating the scope of the essay and including any games at all. But since it was mentioned, you might try some of the newer releases from Loki.
    
    
14. Another large mistake on my part was lamenting the lack of a Lotus Notes client. I'm told that Notes R5 can be accessed from a standard browser. I knew that was the case for administration, but was unaware that this was the case for the client.
    
    
15. One person corrected me for calling WINE an "emulator." WINE actually stands for "WINE Is Not an Emulator". It's a set of libraries that provide Windows compatibility to Linux. I stand corrected. The reason I used the word "emulator" in the first place was because it was easier to explain to corporate department heads. It still is. However, it's important to note that WINE only works on Intel boxes, just like Windows itself. I was also taken to task about WINE's stability. If you want stability you can always use VMWare, but that means that you'll have to have a Windows license, negating some of Linux's unique benefits (low cost, for one). However, it's a terrific way for a user of one OS to run another without rebooting. However, in general, my position is that if you specifically need Windows, you should be running Windows. If you don't specifically need it, you should be considering Linux instead.
    
    
16. Even though the entire section labeled "The Future" is pure opinion, I did get a message from a reader who had a problem with my statement that Windows rode into the Enterprise on the coattails of its home use. Sorry, but that's how it happened in my neck of the woods. Businesses did not jump from DOS and OS/2 to Windows because of Windows' technical superiority, thus forcing it on poor unsuspecting home users. In 1994 I was working at Flagstar Corporation (at the time they were the parent company of Denny's restaurants), which had an all-OS/2 IS shop at the time. Warp and Win95 came out at approximately the same time, and IBM had just dropped the identical business suits in favor of identical Polo shirts... they sent a team out to sell Warp, and Denny's bought it. However, I was in the minority that was betting on Win95's eventual market dominance (one fellow still owes me a case of beer). My argument was that marketing meant far more than technical superiority... a bitter lesson hard-learned by thousands of former Amiga users. Users like me (I was a Win95 beta tester and (later) an MSN beta-tester) brought it from home. As it so happened, the CIO-equivalent (I'm sorry to say I can't remember his actual title) also liked Windows95 at home and we switched over shortly after it came out of beta.

So there you have it. A fairly complete list of the errors in the essay. I'll update this list when I next update the essay. The revised essay can be found at http://www.cratchit.org/dleigh/linux/Linux_in_the_Enterprise.htm.

Correcting these errors simply makes it a better essay, in my opinion. The bottom line is that Linux is still stable, it's still cheap, and it's still ready for the enterprise on both the desktop and as a server.

-- dave.leigh@cratchit.org