BibleGateway.com Verse Of The Day

Tuesday, August 28, 2007

Crazy Women Drivers

We went to a picnic a few weeks back, and my wife got to "drive" a horse-drawn cart. Here she is with our friend Anita driving the horses.....
Yee-haw!

Wednesday, August 08, 2007

A Few Notes on my Load Tester Ruby Script

A few days ago I posted a Ruby script I use to simulate multiple users hitting a web application. I have a few notes regarding that script:
  1. There is a minor bug in it, that doesn't affect functionality so much as the "logging". I used a global var like a dummy and then reference that in the threads.
  2. I also added a 3rd command line param to tell the script how many times each thread should run through the list of URL's.
  3. I also added a sleep() feature so each thread will pause between each run through the list of URL's. Right now that is just a global var instead of a command line param.
I will have to repost the source to get these changes up here. (Actually the *.rb source file is up, I just haven't regenerated the HTML yet.)

On a related note, I wanted to do a better simulation of the types of requests coming through to our prod box. My clever idea, take the actual requests out of Apache's access_log or (in my case) ssl_request_log.

So how to easily grab all those requests to build my file? The format of my ssl_request_log looks a bit like this (some stuff renamed to protect the innocent):

[08/Aug/2007:14:42:23 -0400] 192.237.32.76 SSLv3 RC4-MD5 "GET /appcontext/en/US/CZN2/css/main.css HTTP/1.1" -

Well I just want the URI starting after the GET, so from the Linux/Unix/Cygwin command line do something like this....
$ cat ssl_request_log | cut -d' ' -f 7 2>&1 | tee URIs.txt
Then in vi, I can just put the "http://myserver" portion of the URL at the beginning of each line with a global replace, and I'll have a file with every request, including those for images, stylesheet *.css files, javascript *.js files, etc.

This is much easier than sitting here trying to piece together URL's by hand.

Monday, August 06, 2007

A Great Interview Question

I had to do a few interviews last week, and some of my standard questions were getting a bit stale. So I went out on Google and searched for some new ammo.

I came across this question: "What is the difference between final, finally, and finalize()?"

At first I thought that was too easy, mostly because I knew the answer, and it seemed fairly obvious. Even so, I broke it out in an interview, and the guy nailed the first two, missed the last one.

I was relating that tale to a coworker afterward, and she said, "what was the question he missed?" She nailed the first two, didn't know the last. She's a good programmer; I've worked side by side with her every day for the past few years, and she missed it. That's when I came to the conclusion this is a great question.

Even good programmers don't know everything, and not many Java programmers really have to get into the bowels of memory allocations and garbage collection. This is a great question, not because I want the guy who can answer all 3, but because I want to see how people handle themselves when they don't know the answer. Sometimes that will tell you more about the person than if they just know all the answers.

The first guy I asked this question got the first two 100% correct, then he flat out said he didn't know what finalize was, hadn't run across it, and then he wrote it down in his notebook and said he would have to look it up.

The second guy I asked, he missed it, but didn't seem to know he missed it. For finally, he stammered about a bit and mentioned it was for garbage collection, then switched gears a little and said you usually use it with a try/catch block. Then for finalize, he said he didn't know what it was.

So both missed it, but was one answer better than the other?

In my mind, first guy won. He admitted what he didn't know and showed some level of ambition to look it up. The second guy tried to dazzle me with a little BS for a minute before admitting he didn't know. He didn't write anything down or mention looking it up, he just racked it up as a missed question.

They both went on to answer other technical questions to my satisfaction, but I came away with a slightly difference perspective of each. If they both just nailed every tech question I threw at them, I wouldn't have that added bit of perspective. And how do you know it isn't just "book smarts" -- a lot of those guys that rattle off every new buzzword and argue about architectures and design patterns can't write code to save their lives.


Other Interview Follies:

A lot of funny things do happen when interviewing. Everything I have ever learned about interviewing seemed like common sense, until I met some of these folks:
  • One of my manager's standard questions: "How would you structure your ideal work day?" (e.g. 50% development, 20% support, etc.) One guy told us he would sleep in, get to work late, of course he would stay late also, but his brain just doesn't work in the morning unless he has 10 cups of coffee, so he would rather sleep. Honesty is NOT always the best policy.
  • One lady told us she mostly does front-end work. When I asked her what she was using to build the interfaces (Swing, JSP, JSF, etc.), she told me she mostly does back-end server work. Nice 180, thanks for coming in today!
  • Another guy just got done telling us about all these web applications he wrote, JSP's, servlets, etc. I asked if he was using STRUTS, and he told me he wasn't sure, he might be using it. Then when I asked what application server he was using, he said UNIX and Windows. I tried to clarify, asking him which J2EE server he was using, he looked confused; so I changed my terminology to "servlet and JSP container", he still looked confused; so I made it multiple choice: "Are you using Websphere, Weblogic, Tomcat, Jboss?" His final answer: "Tomcat sounds familiar."

Wednesday, August 01, 2007

HTTP Load Tester Script in Ruby

Here it is, after running my newly acquired Ruby code to HTML converter script, I am finally able to post decent looking Ruby code. Just not directly on Blogger (yet).

Here's the link to my HTTP "Load Runner" script I have been using to debug certain issues (memory leaks, database connection leaks, etc) with a certain un-named inherited application. It's written in Ruby.

For the more hard-core stuff I use JMeter, but this served two purposes:
1. learning more about Ruby
2. having a quick and simple script for small stuff instead of the heavyweight JMeter.


HTML: http://www.frontiernet.net/~nicholson150/http_load_tester.html
Ruby Source: http://www.frontiernet.net/~nicholson150/http_load_tester.rb

Making Code Pretty For The Web

Any time I post code to this stupid blog, I always run into the same issue. How to make it look like readable code? Seems that just putting code inside <PRE> tags doesn't seem to work -- something about the Google / Blogger engine mangles the crap out of the end if it gets "too long".

For Ruby at least, I just found this blog entry that helps out immensely. It's missing a step, as it assumes you have Rio already, but it was easy enough to figure out.
> gem install rio
> gem install syntax
Then save that Ruby script off locally and run it against your script file. Include the CSS or just add it to the HTML inside <STYLE> tags and you're golden.

Of course, the braindead Blogger engine doesn't seem to be picking up the stylesheet, so the code doesn't appear as pretty as it should -- no syntax coloring or anything. I'll have ot figure that out before I post my HTTP load runner script.

Tuesday, July 31, 2007

Using JConsole To Connect To Remote JVM

We have been having issues with a certain Java web application we inherited. Actually, we've had a few problems with it, but that's another story.

One issue was that, even though there is nothing running on this Linux box except a single instance of Tomcat (5.5) with this single web application deployed to it, we will get into frequent issues where we get paged for "high CPU utilization". Log in to the machine, and the Java process is taking 100% CPU. It might stay like that for a few hours, or even a few days. Occasionally it clears up on it's own, but usually the server stops responding and we have to restart Tomcat.

We have had another issue where we would get Out of Memory exceptions (Heap space), and the JVM would stop running. The two issues weren't necessarily related, but we couldn't rule out the possibility either.

We don't have much visibility into the application and what's going on since it is a vendor built "black box". We have some customized source code, but most of it is off limits to us. They rolled their own database connection pool, MVC framework, persistence framework, etc.

In comes Jconsole. JConsole is an awesome tool that comes bundled with JDK 1.5 and above. It connects to the JVM and gives you all the info you could want on the various JVM memory pools, garbage collection, threads, classloading, and lets you manage anything exposed via JMX. It also has lots of pretty graphs, such as memory usage and garbage collections over time, for any or all of the memory pools (heap, non-heap, or individual pools, like permgen and eden space). Same goes for threads and loaded classes --current number, peak, total created.

The best thing about JConsole is the ability to connect to remote JVM's so you don't add too much overhead on the box being monitored. I have JConsole hooked up to my test and production servers, and it helped me prove that the two issues above were connected. There is some condition in the application (yet to be found, actually, first step was proof of what's really happening) that causes a substantial memory leak, and once the memory usage gets at it's ceiling, the garbage collection thread basically runs constantly, trying to regain some trivial amount of memory, then filling it up, and running GC again.

With an average of 100 active sessions at a time plus full garbage collection running non-stop, the CPU gets consumed quickly. I monitored the app for about a week and a half, and memory and CPU looked great -- there were a bit over 200 full GC's in that time period. Then this past weekend, we had to restart because of a DNS issue. That was Friday evening, and by Monday morning there were over 2000 full GC's performed, and I was restarting a non-responsive server by lunch time.

Below are two screenshots, one is of several days of "normal" memory usage, notice gradual rise and then sharp decrease at full GC, all the while keeping well below the JVM's allotted memory ceiling. Second is this past weekend's issue, where memory is hovering at ceiling and a full GC doesn't do much. Also notice the old generation memory pool is quite full.




MAKING IT HAPPEN

To set up JConsole to run on a local JVM, you only need ot pass one extra argument to the JVM:
-Dcom.sun.management.jmxremote
To set up remote (with no security), it is a matter of adding a few more parameters to the remote JVM at startup:
-Dcom.sun.management.jmxremote.port=8004 -Dcom.sun.management.jmxremote.authenticate=false -Dcom.sun.management.jmxremote.ssl=false
Then when you start up JConsole, go to the remote tab, enter the server name and the port that you specified (in this case, 8004).

Simple as that. It starts collecting stats immediately and the graphs appear. As you explore the JMX tree, you will notice yo ucan click on some of the stats and the simple integer displays "opens up" into a full graph display. I'm doing this to watch the active sessions patterns through the day and week.

Monday, July 09, 2007

I Feel So Naughty

Most of my development experience has been with structured languages such as C and Java.

Given that, I feel almost naughty working with Ruby's loose typing. It's like I am getting away with something, doing something oh so wrong. For some reason it never seemed odd with shell scripts, but does here. It's both liberating and a little scary at the same time.

No compiler errors, no warnings, no runtime exceptions, no type mismatches, nothing. You can make x = 3 then turn around and make x = "dude", and then set x = FileRenamer.new with out any issues.

Sunday, July 08, 2007

And The Tree Came Crashing Down

We went to church this morning, and then off to pick up the kids. When we got home, everything was wet, so we knew it had rained. Then for some reason, I felt the need to go over and check on the chickens. What I saw amazed me. 3 big branches of the maple tree had come crashing down on the coop and the fenced in run. I went back to check it out, and the chicken run was completely full of branches and leaves -- I couldn't see the chickens, but I could hear them clucking and running around through the leaves. The back part of the fence was completely down to the ground.

Luckily, no damage was done to the roof of the coop, no chickens were hurt, and once I cleared out all the branches, I was able to repair the fence with minimal effort. Now it's mostly a cleanup effort.

Another Picture of Lexi


Here's an updated picture of Lexi the Wheaten Terrier. She's growing up and getting a lot lighter now. You can see a lot more of the white hair around her face and her belly is almost completely white now.

Friday, July 06, 2007

Black Raspberry Time

It's that time of year again. The black raspberries are starting to ripen in the back yard and in the woods behind the house. The berries seem to be quite a bit smaller than normal, and a little less abundant than normal, probably because we've had a lack of rain so far this summer. But the berries that are out there are quite yummy. We picked a small bowl on the 4th, and I picked a handful today after work. In the morning, I plan on taking my big bowl out again for round two. Normally I can fill a large mixing bowl in one pass, and this time I just got manybe 1/5 of a bowl the first time through.

Tuesday, July 03, 2007

Office Art, Part 1

I drew this on the back of one of my project folders and snapped the picture with my ViewSonic v36 Pocket PC. This is what happens when I get bored in meetings.

Tuesday, June 26, 2007

And 20 Became 19.

fresh chicken and ribs on the barbieUnce upon a time, five Buff Orpington roosters lived among others in a backyard flock, somewhere in upstate New York. A lone Speckled Sussex was even more the outsider, for the flock was mostly made up of Rhode Island Reds.

One sunny day in the summer of 2007, the largest rooster of the flock, a Rhode Island Red who goes by no name, was set aside to become dinner. By "set aside", what I really mean is he was placed into the fenced in backyard, separated from the other chickens and their food, until the time came. But this fenced in yard also happens to be inhabited part-time by a Soft Coated Wheaten Terrier named Lexi.

And so it was, Lexi was let outside and chased the RIR rooster before I could grab either one of them. The rooster ran into some high weeds, the remnants of what used to be a nice flower garden. Lexi ran in after, and came out seconds later, but the rooster did not appear. The chicken must have found escape from the fenced yard, for nobody could find the rooster after much searching.

So the next in line was desperately chosen, as dinner time was drawing nearer with no poultry ready to grill. It was a tough choice, but finally one Buff Orpington rooster stood out as the second largest, and soon succumbed to the meat cleaver. One chop and that's all she wrote.

That evening, dinner plates were stacked high of barbequed pork ribs and a fresh chicken, right from the back yard. Wegman's Memphis Style Barbeque sauce and Frank's Red Hot sauce complimented the smoked meats well.

Soon thereafter, the orginal RIR rooster showed himself. A master of camoflage he proved himself to be, as nobody could find him a few hours earlier, but there he was, right where he had last been spied. He managed to avoid his fate on that June day. But alas, he shall become dinner on another eve, when the 20 will become 18.

Thursday, June 21, 2007

NYS Child Car Seat Inspection

If you're a New Yorker, this is a good deal. It's one of those rare instances where you feel like you're actually getting something back for all the insane taxes you pay. You can go to these free child car seat inspections where officers will inspect your car seats to see if they meet current regulations, are installed properly, and to see that your child fits properly.

If you child has outgrown the seat, or if the seat fails to meet current regulations for any reason, they hand you a new replacement seat. Yes, a FREE replacement seat. They also let you know about the various height and weight guidelines for the different types of seats so you know better what you should be doing. We went to one last month and ended up getting 2 new seats because Raina had just outgrown the seat she was in, and Taryn was very close to outgrowing hers.

Another bonus was free popcorn, hotdogs, and pop. Can't go wrong there even if your seats are up to code.

Check out http://www.nysgtsc.state.ny.us/seat-cal.htm to see when the inspections will be held near you.

Robb's Salsa Recipe

I made a batch of salsa last night, and figured I would share my "recipe", as if there's anything to making salsa -- put crap in food processor and chop the hell out of it.

This is how I did it last night, but it usually varies a bit depending on what's available at Wegmans.

Ingredients
  • 2-3 lbs. "Tomatoes on the Vine"
  • 1-2 Vidalia onions, depending on size
  • 2 -3 Jalapeno peppers
  • 1 big Poblano pepper
  • 1 tbsp. Fresh of chopped Garlic
  • 1 bunch of fresh Cilantro
  • 2 nectarines or peaches
Directions
  • Heat cast iron skillet on high, get it really hot. Add a little dab of oil.
  • Cut poblano pepper and onion into quarters and place in skillet to sear outside. You don't want to completely cook the onions and peppers, just sear them on the outside to bring out a nice roasted smoky flavor and soften them up just a tad.
  • Place everything into a food processor and chop the crap out of it. May need to do multiple batches in the chopper depending on its size. Chop to a consistency you like, some like it chunky, some smooth, and everything in between.
  • Pour into a sealable bowl and store in refrigerator. Let the flavors meld together for awhile before enjoying for best results.
  • If you choose to share, DON'T DOUBLE DIP! That's nasty.
Variations
  • The peaches sound odd or even downright gross if you've never tried it, but they add a certain sweetness without being overbearing. BTW, peaches and nectarines are pretty much the same thing, except peaches have fuzz and nectarines don't. I prefer the non-fuzzy nectarine.
  • Roasting the vegetables on the grill would probably be even tastier than the cast iron skillet method.
  • Add corn and black beans, and maybe even a bit of cumin for a bit of southwestern flair
  • To turn up the heat, add more Jalapenos, Habaneros, or some of your favorite hot sauce. I have young children now, so I try to not get carried away with heat.
  • Of course, green peppers are popular in salsa. I find them a bit overbearing though when raw.

Tuesday, June 19, 2007

Monkey Butt

Sometimes I like to enter random word combos into Google to see what comes up. Tonight I entered "monkey butt" and found an interesting product. The funniest part is, it's a real product.

Thursday, June 14, 2007

Getting Thread Dump Of Running Java Process

Of course you can kill -3 a process to get the thread dump, but for Java apps you can also get thread dumps of running processes using the JDK 5 tools jps and jstack. For example, on a server where I am running Tomcat 5.5, typing jps gets a list of all Java processes running:
-bash-3.00$ jps
25515 Bootstrap
26912 Jps
Then using that process id, run jstack:

jstack 25515

Tuesday, June 12, 2007

My First Vaguely Useful Ruby Script

Alright, anyone with directories full of digital photos, mp3's, or other docs might appreciate an easy way to rename all the files in a directory, either by appending a common prefix, postfix (or extension), or giving each file the same basename with a sequence appending to the filename. There have been times I would have found that useful, and it turns out it's very easy to do with some simple Ruby scripting....


file_renamer.rb


#!/usr/bin/env ruby

require 'fileutils'
require 'find'

class FileRenamer

attr_accessor :directory

# constructor
def initialize(directory=".")
@directory = directory
FileUtils.cd(@directory, :verbose => true)
end

#add prefix to each file in dir, ignoring directories
def add_prefix(prefix)
Dir.foreach(@directory) {|x| rename_file(x,"#{prefix}#{x}") }
end

#add postfix to end of each file in dir, ignoring directories
def add_postfix(postfix)
Dir.foreach(@directory) {|x| rename_file(x,"#{x}#{postfix}") }
end

#rename a file unless it is a directory
def rename_file(from, to)
print "#{from}"
if FileTest.directory?(from)
puts "/ remains unchanged"
else
FileUtils.mv(from, to, :verbose => false)
puts " changed to #{to}"
end #if
end

# rename all files in directory to base_name + a sequence
def rename_with_base(base_name)
idx = 0
Dir.foreach(@directory) do |x|
if !FileTest.directory?(x)
rename_file(x,"#{base_name}#{idx}")
idx += 1
end
end
end

def list_dir
Dir.foreach(@directory) do |x|
print "#{x}"
if FileTest.directory?(x)
puts "/"
elsif FileTest.executable?(x)
puts "*"
elsif FileTest.symlink?(x)
puts "->"
end #if
end #do
end
end

# test it...
fn = FileRenamer.new("C:/test")
fn.list_dir
fn.add_prefix("test")
fn.rename_with_base("start")
fn.add_postfix(".txt")

Monday, June 04, 2007

Set The Captives Free!

My original plan was to let my chickens free-range during the days so they could eat weeds and bugs off my property. I've read that free-ranging chickens will find enough food that you really don't even need to feed them in the summer months. I'm not ready to test that one yet, but I did try a little free ranging this weekend.

On Saturday, I opened up the gate while I went in to get them fresh water, and by the time I got back with the water, they were all out in the yard scratching and pecking and making happy little chicken noises. I let that go on for a few hours while I played outside with the girls.

On Sunday I left them out for a few hours again. Actually I left them out twice Sunday, for a few hours the first time and just about 20 minutes the second time. The second time was a fluke. They used to never try to leave even with the gate open. But now they know what it's like out there, and they leave whenever they can.

I liked seeing the chickens roam about on their own. They mostly stayed right near the coop, foraging in the high weeds for bugs and making the occasional foray into the garden. I still only do it when I am outside with them, though, because some of our neighbors don't feel the need to control their dogs, and just let them roam about in other people's yards.

Lexi got out of the yard at one point Sunday and had a good chase with the chickens. She was within about 4 inches of a Buff Orpington dinner at one point right before I caught her. If a small puppy can be that fast and aggressive toward the chickens, I hate to think what would happen if full grown dogs came into our yard with the chickens out. So I still leave them "ccoped" up when I'm not right there watching and still keep them full of 20% grower/starter and cracked corn.

Wednesday, May 30, 2007

Ruby On Rails On Oracle Hurdles, Part 2

I've hit my first bumps in the road to hooking up RoR to an existing Oracle database schema. Like I posted yesterday, designing an RoR app from the ground up would be easier because you can build your database schema to RoR's expectations and limitations. And now I explain in more detail why...

Other than getting connected to an Oracle database (officially "Hurdle #1"), which I discussed yesterday, I've hit a few more snags.

Hurdle #2: Table naming. I'm trying to hook up to an in-house authentication/authorization database that is used by several of our application platforms. For RoR's "coding by convention" to work properly, you would ideally have plural table names, e.g. reps, rep_roles. Then when you generate your scaffolding, you would use singular names for the controller and view, e.g. rep, rep_role.

The existing database had singular table names. Generating scaffolding would come up with an error about the reps table not existing. Of course not, it's rep. My initial thought is, "how can I fool RoR?" And the answer was obvious, create views and see what happens.


create view reps as select * from rep;
create view rep_roles as select * from rep_role;


With newly created views sitting in from of the tables, I go back to the trusty command line and try to generate my scaffolding again. PRESTO!! It worked. It's a hack, it's trashy, it's shameful. But it works, and it was quicker than ripping apart the framework and touching code in inappropriate fashion.

So, Hurdle #2 has been overcome.

Hurdle #3: This one is what I alluded to yesterday about simple keys. These tables don't have simple "id" type keys. In MySQL when I was starting an app from scratch, of course I would use an auto-increment id. Works great. But according to the RoR on Oracle article (http://www.oracle.com/technology/pub/articles/saternos-rails.html), there are expectations for the primary key column being "id" and having a sequence with a naming convention of "singular-table-name"_seq, e.g. rep_seq for table reps.

Well, that's not going to happen here, not for this database. And despite the tagline of the article ("Learn techniques for creating a Ruby on Rails Web application that utilizes a legacy schema."), the article really doesn't address natural primary keys, whether they are simple one-column keys or complex multiple-column keys.

I can generate the scaffolding and launch the server. I get pages displaying all the rows in the table. But clicking on the show, edit, etc. links fail. The error, obviously is that it couldn't find an "id" column to look up the record by. I also noticed the PK fields aren't showing up on the list view or in the "create new" record view.

I've been working on other projects most of the day, so haven't had a chance to work on it beyond identifying what doesn't work.

So, Hurdle #3 remains a hurdle for now.

And I just had a few minutes to Google, and found this article (http://www.oracle.com/technology/pub/articles/tate-activeerecord.html) which basically says the same thing I am realizing. To quote:
But Active Record is not for every application. It does not handle legacy schemas nearly as well. For those, you need a mapping framework, not one that merely wraps tables with simple classes.

But then there is this blog (http://paulmwatson.com/journal/category/ruby-on-rails/) that has some encouraging words about setting the table name and PK column with the set_primary_key and set_table_name methods of the ActiveRecord model. (The specific post is at http://paulmwatson.com/journal/2006/11/02/wrangling-rails/)

Now that is worth a try, but still not so sure it will work for a VARCHAR primary key field, and still not going to cut it for multiple-column primary keys.

Tuesday, May 29, 2007

Hooking Up Ruby To Oracle

Well, I have to admit this one stumped me for a few days. Actually it stumped me one day, and I got pulled to work on actuall work stuff, and just got back to it a few days later.

I started by following the examples at http://www.oracle.com/technology/pub/articles/haefel-oracle-ruby.html
to get the Ruby Oracle libraries installed. That part was easy, download the OCI8 Ruby package at http://rubyforge.org/projects/ruby-oci8 and run it.

Then I whipped up a script to test the connection from an example I found online (forget the URL to reference -- sorry if I stole your crap, dude!) That script basically looks like this (with usernames and passwords changed to protect the innocent.

test-conn.rb


require 'oci8'

OCI8.new("user", "password", "oracle_sid").exec("select * from rep") do
|r| puts r.join(", ");
end



I was pumped! That seems easy, and then I can get out of the MySQL realm and start hooking up Rails to existing enterprise DB's and really show this stuff to be useful in our environment. Sorry MySQL fanatics, we are (mostly) an Oracle shop -- I have nothing against MySQL for my personal projects.

Anyways, she didn't work out so well. I figured it would work since I have Oracle installed on my PC, but got the following popup when I ran that script.

There was also a Ruby stack trace on the command line. Basically I did some Google'ing and found this posting (http://blade.nagaokaut.ac.jp/cgi-bin/scat.rb/ruby/ruby-talk/238003) which at least gave me the clue I needed to look for OCI.dll on my system. My thought was it was getting confused between the Oracle 8 and Oracle 9 versions loaded on my machine.

Searching and searching, I don't have OCI.dll on my machine. Back to Google (a developer's best friend IMO, it always confuses me when other developers don't try this first when they get stumped) That's when I came across this page (http://www.dll-files.com/dllindex/dll-files.shtml?oci) that let me download OCI.dll. Dropped that in C:\WINNT\system32 (I'm on Windows 2000 at work, might be WINDOWS instead of WINNT for you), and re-ran the above test-conn.rb script.

IT WORKS!
Ruby on Rails on Oracle, here I come (again)

I'm guessing my next big hurdle (or potentially a roadblock instead of a hurdle) will be the lack of support for complex primary keys. Starting an application from scratch, you can design your database with that limitation in mind, but building applications for existing databases becomes another issue. I'm sure someone smarter than me has already addressed these concerns, so it might just be a matter of more Google searches.