Mark Mischke's Weblog

Separation of IT engineering from support

In the brief time between when I left the field of Chemistry and when I formally got into software development, I ventured into the world of Information Technology. The term IT technically encompasses everything to do with computers. When used in the common vernacular, it tends to refer to the network, server and desktop computer infrastructure, in an architectural, maintenance and support sense.

What intrigued me about IT was the technology, and how it all fits together. It's every bit as much of an engineering challenge as developing software is. Maybe even more so, since the spectrum is broader. I got into IT back in '92, just as Windows NT was coming out of beta. I became engrossed in all things NT, and I wound up leading several corporate infrastructure migrations from Netware and VMS to NT. I had a background in C/DOS/Win16/Win32 programming, and it always felt like custom development (especially scripting) was a natural aspect of server administration. Strangely enough, while this is true among the UNIX IT crowd, it seems to escape many MS IT admins.

The thing that ultimately drove me away was the support aspect. Once NT started taking root, my role started shifting from a purely engineering role to a combined engineering/support role. So I bailed. I couldn't understand why the engineering and support functions were so often blended. They're diametrically opposite, in terms of skill set and personal disposition. I just don't have the empathy to do support. When someone says "I can't print", my skin crawls. :-) The first thought that crosses my mind is that the company dropped a couple grand on their computer and they've taken no responsibility for learning even the very basics. I know plenty of non-techies who have made the computer a partner, and not an adversary, so I feel that most excuses are weak.

I still love doing IT stuff, but I don't let too many people know it, because the next thing you know, somebody can't print. :-) The IT background helps me immensely in my role as a software developer, and also helps me win the trust of the IT folks, which is essential to the success of many software projects.

IT in America is, almost by definition, a reactionary function, despite the fact that it serves a mission-critical role. Several years ago, I heard the term 'data janitor' used to describe the IT professional. Truer words were never spoken. IT has to clean up every mess that users (and developers) create. An employee will get zinged if he expenses a hooker. Why doesn't he get similarly zinged if he sucks up valuable company resources by having IT configure his printer? Say what you may, but I expect everyone in corporate America to understand the very basics of computing. I expect you to understand the concept of folders and files. I expect you to know how to set up your printers, providing IT publishes good instructions. I expect you to know that the T: drive is on the network, and not inside your computer. That's not really asking very much, considering the investment the company has made in the individual.

IT has two distinct functions - strategic engineering and user support. The two are quite different, and require very different kinds of mindsets. The guy who built your infrastructure doesn't give a rat's ass if you can't print. He's more concerned with the odd TCP port requests that he saw in the firewall logs this morning. He's an engineer, and doesn't want to help you, nor should he. There should be an empathic, concerned IT help desk technician for that. He cares that you can't print, and he wants to help you get squared away.

Admins are spread very thin. We expect them to keep our mail servers operating as reliably as the telephones and the electricity. We pay them as little as we can, give them a meager budget for hardware and expect them to cover every other task that falls into the IT bucket - AND keep the friggin' e-mail flowing. The telephone and electrical infrastructures are supported by billion-dollar industries. We don't even even separate the IT engineering function from the support function. That's like making the CEO also assume the HR role, or making the Marketing department do sales.

.Net for the masses

Microsoft has positioned its .Net Application Framework as the present and future of Windows development, whether it be on the desktop, the Web or under the hood. This revolution began abruptly and absolutely over two years ago. As we speak, Windows 2003 Server is the only Microsoft operating system to have native support for .Net applications.

I'm perplexed as to why Windows 2000, and especially XP, haven't yet been mandatorily retrofitted with .Net support. It's an optional add-on via Windows Update, but most users don't know or care what .Net is. They just want to be able to run applications.

The .Net runtime (only 21 MB in size) should have been part of every Win2k and XP service pack that was released since .Net went gold. Service packs are well over 100 MB as it is, so what's another twenty percent, especially if it represents what Microsoft considers core operating system functionality. If developers can't count on .Net being on users' machines, they're going to hesitate when it comes to creating .Net apps. It's not fair to expect the user to download and install the Framework himself, nor is it acceptable to expect the developer to bundle it with his app. .Net apps tend to be very small in size, with most of the heavy lifting being done by the Framework. A vendor isn't going to bundle a 21 MB installer along with thier 700k app.

Microsoft has aggressively pushed .Net from the very beginning. When they said that .Net was the future, and that pre-existing technologies were (essentially) deprecated, every MS developer I knew dropped VB, ASP (etc.) like hot potatos and got on the .Net bandwagon.

It's now two years later, and I, for one, would like to be able to count on the .Net Framework already being installed on users' systems. I figure it must have to do with the resources required by .Net. There are probably a lot of machines out there that can't effectively host it. MS has some very smart people. Surely they can find a way to ensure that .Net finds its way onto every box that can support it.

Why is Windows 2003 Server so damned secure out of the box???

To all you freaks who whined about NT/2k being too insecure - you suck! I just upgraded a 2k Server installation to 2003 Server and now nothing works. I have to undo all kinds of security settings just because too many NT/2k admins don't know their ass from a hole in the wall, and never bothered to secure thier servers, forcing Microsoft to do it with 2003.

A few years back, I worked for one of the largest ISPs in the world, and one of my jobs was to make sure that NT boxes could sit out on the 'Net, unfirewalled, and take bullets all day. It's not that hard, folks. Despite what the /. crowd might say, my boxes were never compromised, and NT is every bit as secure as whatever you might compare it to - if you secure it.

I'll stop ranting now. I've got a server to reconfigure. I'm just a little pissed because the upgrade process from 2k (yes, I broke my own cardinal rule - never upgrade OSs) tightened stuff down so much that nothing works. :-)

Why isn't spyware illegal?

Would people tolerate their toasters secretly turning themselves on at night and consuming electricity? What if TVs changed the channel to Yan Can Cook in the middle of The Simpsons?

Why isn't adware and spyware illegal? This software changes the way your PC operates, regardless of whether you're paying for your bandwidth, or not. Even worse, it tries to hide itself.

What if your toaster was rated as drawing 500 watts, but when you weren't looking (or toasting) it drew 5000 watts, just for kicks?

Enough is enough. Let's send these adware/spyware jerks to jail, just like we would if someone tried to sell a smoke detector that transmitted audio feeds back to the manufacturer. It all feels very similar to a breaking-and-entering in which the perpetrator takes notes on your life and sits in your La-Z-Boy and watches your large-screen TV while you're not home.

If computing platforms were women...

...then:

Apple Woman would be perfect in every way. She would be beautiful, completely devoid of bad habits and an ideal partner. She would never expect or want fancy jewelry, and she wouldn't mind if you forgot her birthday or your anniversary. Her sole purpose would be to anticipate your needs and please you in every way possible. She'd be a Stepford Wife.

Microsoft Woman would be very much like the girl next door. She'd be reasonably attractive, and would become more so as you got to know her. She would aim to please you and to be a good partner, but she'd also make you work a little bit in order to keep her. The longer you're with her, the more she'd understand you and she'd reward your devotion by becoming more and more attentive.

Linux Woman ships looking like a crack 'ho. If you're really nice to her, she might put on some makeup for you once in awhile, but usually only on one side of her face. She comes with plastic surgery and psychotherapy modules, allowing you to fine-tune her looks and her attitude to suit your needs. The documentation, assuming that you can find it (the link included with the kit is dead), is written in Cyrillic by a native Spanish speaker with no knowledge of Russian. Screenshots abound on the Internet, showing how Linux Woman can be made into the perfect Stepford Wife. After twelve hours of recompiling her attitude, you'd be happy to just get her to stop yelling at you. This is the point at which most men uninstall Linux Woman and reinstall Microsoft Woman.

Amiga Woman was a knockout back in the day, but the years haven't been kind to her. She requires a man who will defend her honor at the drop of a hat. She expects you to start a fight if you catch anyone NOT checking her out, regardless of whether the observer is single, married, gay or female.

OK, now before anyone gets pissed at me for being a sexist pig, keep in mind that similar definitions could easily be crafted for men. Howzabout someone take the baton and send me a link?...

Customer service as part of the user experience design

My poor iBook. It got sick again last week. It would randomly freeze, boot to a black screen or vomit colorful distorted patterns all over the place. I did some research and discovered that Apple had issued a recall for this problem - a fault in the logic board in which the graphics chip can become loose.

My first problem with the machine happened when I'd only had it for about three months. The display would go dark if the lid angle was changed. I discovered that iBooks have a poor display umbilical design, which frays easily. I called Apple, and was blown away by the experience. I got no runaround at all. The tech sent me a special padded box overnight, complete with a pre-paid shipping label. I sent it off, they fixed it the next day and three days after my call, I had the working iBook back in my hands. The whole process was unexpectedly pleasant and painless.

I experienced the same thing with my recent problem. Since the machine was well out of its warranty period, they made me do a few tests over the phone to determine that I did, in fact, have the problem addressed by the recall. After that, they sent me a shipping crate, and my iBook was back home in three days.

Somebody at Apple seems to understand that usability and the user experience doesn't stop at the hardware or the software. Apple is a company that's all about usability. Thier computers are designed to work the way humans do, from the metal to the OS. When one of those things stops working as it should, Apple recognizes that the process of fixing the problem should be as user friendly as its Aqua UI.

So, here I sit with an iBook that's been to the hospital twice in its eighteen month life (and I'm VERY easy on it). One might think that would make me hesitant to buy another Apple product in the future. It's quite the opposite. Apple has shown me that they're a partner, and that the partnership continues after the sale is completed.

Linux vs. getting real work done

I've been living with SuSE 9.0 (w/KDE 3.21 shoehorned on) on my HP Pavilion ze1115 notebook for the past couple of weeks. I mainly use it while working at client sites, but I've also been using it as my primary e-mail machine in order to get a good feel for KMail. Workwise, I make heavy use of the rdesktop app to Remote Desktop (it's become a verb) into my main dev box at home, from wherever I might be.

Linux (or at least SuSE 9.0) isn't ready for notebooks yet. If I want to switch between my trackpad and a USB mouse, I have to reconfigure my hardware through YaST. Each client site has a wireless LAN, as does my home. I have to run a specific script for each site, and similarly to my mouse woes, I have to use YaST to choose between my wireless PC Card or the internal Ethernet adapter. If I leave the Ethernet adapter enabled, it steals the default route (which is null if it isn't connected and doesn't get DHCP information). I could fix that with more scripting, but c'mon...

Battery life is significantly reduced when running Linux. I get almost three hours with XP, and barely two with Linux. This isn't that big a deal, because I'm usually plugged in. SuSE backported some power management features of the 2.6 kernel into thier 2.4 kernel, but I'm guessing that it doesn't support CPU throttling. I tried installing 2.6 kernels (both RPMs and from source compilation), but each time it was a tradeoff between getting stuff I wanted and losing stuff I had before. I'm sure I could have eventually figured it out, if I had the time to invest.

My understanding is that all of these features will be addressed in the upcoming SuSE 9.1 (with the 2.6 kernel).

For now, I've got XP back on the machine. I have to get real work done, and I can't have my colleagues waiting for me to finish dicking around trying to set up my network connection and my mouse before we can get down to business.

Plus, there's no VPN (meaning PPTP) support. I'm sure there's some free component in an alpha state up on sourceforge with no documentation that I could download, compile, futz with for three hours and still not get to work. OTOH, my iBook supports PPTP, and it's BSD, so it might not be as bleak as I'm portraying it. All I know is that I'll be needing PPTP and I don't have a lot of time to go on a 'Net safari tracking down the solution.

I intend to give SuSE 9.1 a go-round when it's released. I may even get really geeky at some point and build a Gentoo box. Not right now, though. Time to get back to work.

Mischke's First Commandment of Medical Diagnosis

In early 2003 I took a spill down some icy steps, because I had my head up my ass and wasn't watching where I was going. The whole thing was pretty embarrassing, once I determined that I was still in one piece. I like ice. I like running at nearly full speed on icy paths with my Yaktrax. I like cramponing up a frozen peak, sticking like glue the whole time. I also consider myself pretty good at negotiating icy pavement in regular shoes. Needless to say, the experience was humbling.

A month later, my upper arm still hurt when I moved it in certain ways. I eventually made an appointment to see the doc. My doctor's a good dude. He's quick to order the right tests and make referrals to specialists. That suits me fine, because I don't like it when something goes wrong with me, and I like to bring all guns to bear on the problem and make it go away quickly.

I had an MRI and was told that I had two torn tendons and a loose bone fragment, and that I'd need arthroscopic surgery. That was a bummer because it meant that I had to triage my planned Summer activities, because the procedure would keep me from playing in the hills or paddling for a good six weeks. My doc referred me to a surgeon and ordered another MRI, this time with a contrast agent injected into my shoulder to get better images (it didn't hurt a bit).

I went to see the surgeon - a fantastic guy - who looked at the films and said there were no mechanical problems, and that I just needed some physical therapy. I was dumbfounded. Pleased, to say the least, but dumbfounded. WTF was the first diagnosis about??? One guy says I'm a mechanical mess and the other says it's all good. My doc isn't to blame for the initial reading. He just got the results from the hospital, and it's doubtful that he ever saw the films himself. I feel that he did the right thing by taking steps to get me all fixed up.

This still leaves the question of the misread films. All I can figure is that they pulled a guy off the street to read 'em. Either that, or it was some Radiology resident who hadn't slept in thirty-six hours.

Since the problem was just an imbalanced muscle group, I changed the types of weight exercises that I was doing to cover the ones that were being neglected. A month later, I was my old self again.

All this has taught me one thing, which I've dubbed Mischke's First Commandment of Medical Diagnosis:

If Thou Disliketh Thy Diagnosis, Get Thee Another And Another Until Thou Art Pleased.

Let's get it straight

It's Web site, not website.
It's e-mail, not email.
It's Internet, not internet.
A sentence beginning with e-mail starts as: e-Mail.

...And there's no such thing as a webinar...

Despite this, I prefer 'email' to 'e-mail', but I'm still wrong. :-)

Defensive coding vs. excessive exception handling

I was recently wondering if all those omnipresent try/catch blocks create additional overhead or impose a performance hit. I found a plethora of articles and comments that suggested that, at least in the .Net world, there is essentially no penalty for using them. Perhaps that's why so much .Net code is clutterred up with exception handling.

I believe that it's only prudent to add exception-handling code in situations where you might reasonably expect to generate an exception. All to often, try blocks are just convenient containers for sloppy or untested code.

Some example code I've seen (IIS 4.0-era metabase scripts come to mind) actually used exception handling (or what passed for it in VBScript) as a way of testing the results of operations. The code would just puke its way down the ladder until it found the place where it could stop puking. This is a rather extreme example, but some developers do use exception handling in this manner.

I started out as a C guy, and C has no concept of exception handling. It will let you shoot yourself in the foot and then laugh at you for doing it. C developers quickly learn to code defensively against the dreaded null pointers and orphaned data structures. Whenever possible, I avoid adding any exception handling until the application has been thoroughly unit tested. I'm a big fan of defensive coding, even though it's not as critical now as it was with C. Check those objects for null references before using them. Make sure your DataTable or collection has elements before attempting to reference them. Sounds simple, but it's often overlooked. I prefer to bulletproof my code and prevent exceptions from occurring, wherever possible. Then, when it's time for QA, I'll integrate some exception handling where it might be needed. IMHO, ninety percent of all the exception handling out there could be eliminated, making the code much more readable. Good code doesn't need to rely on a black hole to suck up the problems.