Fog Creek Software
Discussion Board

10/90 - code written by 10% is in use by 90%

MS Word, MS Windows, Oracle, CityDesk...

10% of programmers write software which is in use by 90% of users.

But still in every larger company is own, in-house, system, which reflects exactly what the business does. Those systems is about what the rest of programmers or 90% of us is working at (each such system is used may be only by 0.0001% of all computer users).

Everybody knows statistics for internet browser usage, for operating systems usage, but how can I find business software development tools usage statistics? I mean programming languages and tools which may be not providing best user interface or software speed, but allow a team of 5 juniors to build entire in-house system. May be even put it into web.

We use Progress and WebSpeed here. What do you use?

Thursday, October 10, 2002

Interesting point, but I'm sure your stats are way off. I can't think of any program used by 90% of computer users except Windows - and even then only if you lump all the versions together. My bet is even Word doesn't have 90% installation, let alone 90% usage.

So the only developers writing for a 90% audience are the Windows OS team and the freebies that come with it- surely much less than 10% of developers. 0.001%?
0.00001? less?

David Clayworth
Thursday, October 10, 2002

David, you're right. But I didn't mean exact stats. Let me try to make myself clear: I just want to figure out, that people outside my office use for rapid development of in-house systems. When much more important system development speed and cost, than portability and usability. When you have just average or let me say below average technicians who much more concerned about their kids and beer, than computers and technologies. But you still need to accomplish task and deliver "good enough" system on time, to let business function.

So, what do you use: Oracle and pl/sql? Java? VB? Delphi?

Would be nice if everyone could mention tools they really use today, developing in-house.

Thursday, October 10, 2002

VB is really common for lowest-common-denominator projects.

No offense to anyone, but VB fufills the old requirements:

1) Runs good on windoz
2) Easy to learn
3) Fast to develop
4) Who cares about execution speed? It's an internal app!
5) Cheap-O compiler

Delphi is like VB, only a bit harder to learn and, if you built it "right", a good bit easier to maintain and extend.

In other words:

A certain kind of Smarty uses Delphi.

VB is great, it's neat, it's fast, it's effective.  However, it's _SO_ easy to use that it attracts a certain type of person  ... and heaven help you if you are the Real Programmer (TM) brought in to clean up the mess made by the Accidental Programmer (TM).


Matt H.
Thursday, October 10, 2002

if development speed & cost are more important than portability & usability, then I'm afraid you're going to find it awfully difficult to keep these "below average technicians" up and running.  In my experience (YMMV), the developers are the ones able to work better with less usable tools, and the technicians are the first to admit defeat over a complex interface or puzzling command line syntax.  If you don't pay some good attention to usability, even on in-house systems, your developers will spend a lot of time answering questions, and not developing.

As to your clarified question, for in-house systems we use java and scripting languages (perl, tcl/tk mostly).  And as we have a massive corporate intranet, we do a lot of CGI systems, usually in perl.

van pelt
Thursday, October 10, 2002

This is totally off-topic, but whenever someone comes up with a stat like this, why is it that the numbers always sum to 100%?  Like the 80/20 rule (80% of execution time is spent in 20% of the code, sometimes incompatibly called the 90/10 rule) and countless others.  The people who create these rules of thumb know that the percents refer to 2 different things, and so don't have to sum to 100%, right?

Thursday, October 10, 2002

VB/VBA and the term rapid application development have tended to go hand-in-hand over the years. 

Perl would be another obvious candidate.

Charles Kiley
Thursday, October 10, 2002

Yo Vlad,

"below average technicians who much more concerned about their kids and beer, than computers and technologies"

Hey, I noticed this little swipe you hid there in your post.

You really think so? A programmer that cares more about his kids than computers and technology is 'below average'? In my experience, I'd say there's a little correlation there, but it swings the opposite way.

And as far as the beer goes, I don't trust a programmer that doesn't understand the value of a good beer. Look at the Germans -- the world's finest engineers as well as hearty advocates of fine beer.

Don't forget -- Beer makes you strong. The Babylonians knew it -- their top engineers were all paid in beer and they made some mighty fine ziggaurauts.

Ed the Millwright
Thursday, October 10, 2002

Vlad wrote:

"below average technicians who much more concerned about their kids and beer, than computers and technologies."

So a person who cares more about his family than technology is somehow "below average"?

Wow. That's a pretty strong statement there.  I'd say someone who doesn't live up to the responsibilities that he owes his family is the one that needs help.

I forget the title, but a former Microsoftie wrote a book about the death march to ship Windows NT 4.0 and how Redmond was littered with broken marriages, adultery and children that spent 2-3 years without one of their parents all in the name of shipping NT and climbing the corporate ladder.

While I have a huge distate for "Cowboy Coders" and their ilk, I have a deeper suspicion of those people that are all to ready to sacrifice their families for the sake of success.

Mark Hoffman
Thursday, October 10, 2002

Yes, and what about us single wine drinkers?

Are we also bad coders?

Thursday, October 10, 2002

The book is "Showstoppers" (about NT).  It's great!

Marcus Blankenship
Thursday, October 10, 2002

I have usually used Java/MySQL, developed under something like Eclipse.

Friday, October 11, 2002

ha-ha-ha, my apologizes to you all beer, wine, whisky, vodka, sake, schnapps, puskar, tequila, samogon, pertsowka and cognac lovers! I do not undervalue alcohol or our love ones.

Lets go back to the topic. Thank you for responses, they do make sense for me.

Friday, October 11, 2002

  You forgot to mention Cachaça.

Ricardo Antunes da Costa
Friday, October 11, 2002

  And getting back to the topic, I used to use Clarion.

Ricardo Antunes da Costa
Friday, October 11, 2002


not many people in this forum will understand what is "samogon" :)

Friday, October 11, 2002

btw, the book "Showstoppers" was about the development of Windows NT 3.5 (aka NT 1.0), not NT 4.0.

Zwarm Monkey
Friday, October 11, 2002

...and what about kvass?  OK, it isn't exactly alcoholic, but it's worth mentioning...  : )

van pelt
Friday, October 11, 2002

[Brian] Whenever someone comes up with a stat like this, why is it that the numbers always sum to 100%?

For *any* distribution, it's possible to come up with a statistic that says that the top x% of the range is accounted for by only 100-x% of the domain. Examples include "90% of the energy consumption on earth is due to 10% of the people", "80% of the bugs are in 20% of the code", etc.

Making the numbers add up to 100% merely reduces 1 variable that you need to think about. You can tell that a 90-10 is more skewed than an 80-20, but what's the comparison between an 80-20 and a 90-30?

It should not be a surprize that the first number is bigger than the second -- this is *always* true for any distribution. The only information here is that if the first number is close to 50% then the distribution is nearly uniform, and if it's close to 100% then the distribution is highly skewed.

Unfortunately, this is also an area where people feel free to make up statistics out of thin air, so you need to take these figures with a large grain of salt. (Yes, I made up the two figures above. I suspect that the original poster made up the 90% figure.)

Jim Lyon
Friday, October 11, 2002

>Making the numbers add up to 100% merely reduces 1 variable that you need to think about.

No, it only masks it.  That is his (valid) point.

looking deeply into shallow waters
Friday, October 11, 2002

I'm embarrassed to admit that I didn't realize that there is always an x/100-x point for any distribution.  But I see now that that's the case.

Now back to your regularly schedules discussion.

Saturday, October 12, 2002

*  Recent Topics

*  Fog Creek Home