Troubleshooters.Com Presents

Troubleshooting Professional Magazine

Volume 3, Issue 8, August 1999
Self-Documenting Code
Copyright (C) 1999 by Steve Litt. All rights reserved. Materials from guest authors copyrighted by them and licensed for perpetual use to Troubleshooting Professional Magazine. All rights reserved to the copyright holder, except for items specifically marked otherwise (certain free software source code, GNU/GPL, etc.). All material herein provided "As-Is". User assumes all risk and responsibility for any outcome.

[ Troubleshooters.Com | Back Issues ]


Editors Desk

By Steve Litt
I heard from an old friend the other day. He's a programmer now. Seems these days every technology oriented person is pressed into duty whomping out code. Will it be good code? How is good code defined? Speed? Reliability? Small footprint? Reusability? Readability? Yes.

Let's talk about readability. I'm often brought in to help make the enhance or rewrite decision. The big factor -- readability. The client doesn't want to pay mega-hours just to decide which way to go. So a timely decision must be made, based on the existing code and the needed enhancements and bug fixes. If I can't quickly read the code well enough to understand the original programmer's design intent, I'll probably recommend rewrite and comment on the lack of that programmer's  readability.

In other words, a program can be very well designed from the machine point of view, and even from the human point of view. But if it does not clearly display that design, its usefullness is severely curtailed.

This issue of Troubleshooting Professional Magazine is devoted to code readability. With an ever greater proportion of functionality implemented in software, we Troubleshooters deserve readable code. So kick back, relax, and enjoy this issue of Troubleshooting Professional. Learn my thoughts on readability, and enhance them with your own habits and discoveries. And remember, if you're a Troubleshooter, Technologist, or free software user, this is your magazine. Enjoy!

Steve Litt can be reached at Steve Litt's email address.

Code Readability Throughout the Ages

By Steve Litt
Well, at least throughout the ages when I was a programmer. So lets start with 1982. Culture Club topped the charts, I studied programming at Santa Monica College (SMC), the country was in a the grips of a short but nasty recession, and thousands of spaghetti programmers were being shown the door as their era ended.

I was taught the new methods from the start. Structured programming. Use comments often. Use modern languages like C whose compilers don't care what line or column the code appears in. Our instructors laughed at indentation dependent languages like RPG. Those were the way of the spaghetti specialist, not young turks like us.

Our instructors were split on the subject of Cobol. Yes, it had (at least in those days) no local variables, and it was indentation dependent, but it was "self documenting". What could be clearer than "move employee record to print line."? The school's elite looked down their nose at Cobol, that sissy language. I wasn't so sure. Self-documentation sure had its advantages. I became a professional C programmer in 1984 -- one of SMC's first students to do so. And the more I did C, the more I yearned for readability. I couldn't even work on stuff I'd written six months ago. But that's getting ahead of the story.

Commenting on My G-G-Generation

Above all, our instructors taught us to comment, comment, comment. The more comments the merrier. Comments every 5 lines were just fine. A thesis at the top of each subroutine was ideal. And of course every variable needed its own comment to the right of its declaration. We swaggered out of SMC into the job market, cocksure of our new methodologies. We were the new wave -- we knew it all.

And we rewrote, rewrote, rewrote. All that creeping crud made by the spaghetti swarm, we couldn't read it -- we replaced it. And because our design methodologies were much more scalable than spaghetti producing flow charts, we got the job done quickly. But somehow, our stuff wasn't much more readable than that of the flow chart fellows in the unemployment lines. My reaction: Comment even more. The late 80's and early 90's saw me producing programs that were 30% comments. I never saw the problem. After all, I was jumping around so much I didn't maintain my own code. That changed in the early 90's.

When you maintain commented code, you must change the code and the comments to match each other. Well actually, "must" is too strong a word. In hot deadline situations, many changed the code and left the comments alone. So when the next maintenance programmer came along, he was misled. Kludges followed kludges. Good software gone bad. I started commenting less.

There is None as Blind as He Who Will Not See

Come 1987 I was a manager. I reviewed the code of a junior programmer with a freshly minted BS in Computer Science degree . I was struck by the program's beauty and readability. The main routine was four subroutine calls, each of which was structured similar to the main routine. Written in Pascal, it read like English. But it had taken him forever to write the program -- that must be what happens when you go overboard on modularity and functional decomposition. And there was the efficiency problem.

Sure, we displaced the spaghetti guys in the marketplace, but not before they subtly indoctrinated us. That indoctrination -- write efficient code. Oh sure, we knew to trade some efficiency to substitute subroutine calls for gotos, but we didn't want too many levels of subroutine calls. There's speed and stack to consider.

I told the junior programmer he'd need to be less a artist and more productionist and efficiency expert if he wanted to make it in our shop. I soon moved on, never again seeing the junior programmer or his beautifully crafted code, which I (wrongly) associated with non-productivity. It would be years before I really appreciated what that programmer had done.

You Never Talk To Me Any More

As I maintained more and more of my super-commented code, My mistake became obvious. No matter how hard I tried, code and comment went out of sync. First just a little, then a lot, then enough to cause departure from the intended design. So I commented less. Much less. And I tried to make the code itself do the talking.

Hungary for that Readability

A co-worker introduced me to Hungarian variable naming. Wow! Finally a self-documenting system. First since Cobol. I jumped on the Hungarian bandwagon. For those lucky enough to avoid Microsoft code (Hungarian was named after a Microsoft programmer originating from Hungary), here's the executive summary. Every variable name has a prefix, telling: There were other details I never knew or long since forgot. You could make up your own types for structs you created, or for other purposes. The idea was you'd never again unwittingly pass an int to a routine requiring an string. If you want to see the ultimate in Hungarian naming, look in any documentation of the Windows API. Every var is prepended with stuff like "lpstr".

Hungarian was nice while I was in that shop, because everyone was using it. But I used it less and less after leaving that shop because it didn't deliver the promised readability. And you shouldn't use global vars (g) anyway. And the world was turning away from strongly typed languages...

The Snake You Hate To Love

I hated Python the minute I saw it. Like the RPG my now-retired instructors had snickered at, it was indentation dependent. Instead of using begin and end braces to delineate statements subservient to if, else, while and the like, it used indentation. Yuck!

I'd heard such great things about Python -- why had Guido Van Rossem given it this syntax straight out of the Disco era? But I wanted to see what all my friends were raving about, so I took a few hours coding Python. And after 3 hours I realized, "Man, that's readable!".

Remember all those brace placement feuds that made Windows/Linux look like a friendly chat? Gone! Remember that turkey maintenance programmer who screwed up the indentation, followed by yet another turkey maintenance programmer who believed the design to be represented by the indentation, moved the braces and messed up the program? Gone. With the "compiler" enforcing indentation, what you see is what you get. (and where do you want to go today?:-).

If some foolish maintenance programmer messes up your Python indentation, chances are it will immediately error out, giving a line number. But even if he's unlucky enough to have exdented the final subordinate statement, the wrong answers will quickly let him know there's a problem, and he'll go back to the last area he worked on and find it. And if you're *really* concerned that someone will delete an indent and create a hard to find problem, a single comment saying

    # end of if x < 1
will alert all but the densest maintainers to what's going on. And if you *ULTIMATELY* want to enforce the intended indentation, have the final statement be a call to thisIsTheEndOfTheBlock(), a do-nothing routine taking one argument -- an argument declared in the block that will abort with a NameError out if not earlier declared within the block :-)

To Know You is to Love You

Within four hours I knew Python was the language I had always hoped for. I love it for its flaws like weak type checking and indentation sensitivity. But that's not really on topic.

It's the Readability, Stupid!

What Python taught me is that the best documentation is that enforced by the compiler. I took a long walk in the Florida sunshine, contemplating how that fact could be extended. Variables. Name them well and they tell a story. And if they're named right, the maintenance programmer will use them, instead of inventing parallel ones. What about subroutine names? They're compiler enforcible (at least matching the call to the declaration). Use  enough of them, and the entire program becomes a comment. A COMPILER ENFORCED COMMENT.

Just like the beautiful program the junior programmer showed me so many years before. The one I criticized him for. The one where the guy had subroutines just to execute one statement. Nahh, I couldn't advocate that, could I?

This Ain't The Summer Of Love

The hardest habits to break are the ones that once made sense. I would never make a subroutine to execute a single statement. The spaghetti chefs were before my time, but I worked with enough of them to subconsciously pick up some of their habits. One habit was "count those machine cycles, watch that stack". Made a lot of sense when Gracie Slick sang free in Golden Gate Park.  But it makes absolutely no sense today. Twelve levels of subroutine calls may be a little inefficient, but it's downright svelte compared to the corpulance of the average "development environment" produced program.

Imagine a program taking a person's first and last names as the second and third command line arguments. Which statement is easier reading?

print sys.argv[3]
print getLastname()
Of course, to use the latter there must be code like this somewhere:
def getLastname():
   return sys.argv[3]
But by the time you get to that point, the source of the last name is a mere implementation detail. The point is, the programmer has documented his design intent.

But at what cost? Golly gee, the machine cycles, the stack!

If you're using Perl or Python, chances are performance isn't your priority anyway, so who cares. Performance people like C and C++. And in those languages you can make getLastname() a macro. Have your cake and eat it too. It compiles to argv[3], but the source reads getLastname().

This really starts gaining power when using objects to represent the state of a thing:

print employee.getLastname(), employee.getEmployeeNumber()
print lawyer.getLastname(), lawyer.getPhone(), lawyer.getDegree()
print defendent.getLastname(), defendent.getLastConvictionDate()

But Wait, There's More

And if you order this self-documentation methodology today, you get, at no extra cost:

A design methodology!

That's right. If you order our compiler enforced self documentation methodology, we'll throw in, absolutely free, a design methodology. Want a simple example?

We all know you should never sit down and just start typing a program. Right? You diagram it, or use an outline processor, or use case tools, or whatever. But what if your code looks like English:


def main():
   (age,firstname,lastname,gender) = getCommandLineArguments()
   if correctDemographics(age,gender):

def getCommandLineArguments():
   return (eval(sys.argv[1]), sys.argv[2], sys.argv[3], sys.argv[4])

def correctDemographics(age, gender):
   if  20 <= age < 30 and gender == "male":

def sendSalesLetter(age,firstname,lastname,gender):
   printSalutation(gender, lastname)
   printBody(gender, firstname, lastname)
Of course you need to code the three routines called in sendSalesLetter, and at the very bottom you need a call to main(). But you get the picture. To a certain extent you can type in your design document, which has English like readability. The design doc is also your code. But wait. What about  OOP?   Same thing. Each object is an actor in a script, so name all the leading men and ladies in the main routine. Name them descriptive names -- job titles if you will. Each of their method names exactly describes a single behavior. Then the main routine becomes a script narrating how they all interact with each other. Hardly any syntax -- anyone can understand it, and everyone knows the design intent. Supporting actors might be similarly declared in the bodies or methods of the main actors, and likewise their interactions are narrated, and so forth.

How is This Different From the Self Documentation of Hungarian or Cobol?

First, this method of self-documentation will work beautifully with Cobol programs, and it will work just fine with Hungarian variables too, as long as the programmer doesn't skimp on descriptive names to compensate for the prefix typing.

Let's look at Cobol. It's self-documenting at a computer instruction level, but not *necessarily* at a conceptual level. That wordiness:

move employee-record to printline
is really not a whole lot more descriptive, for somebody knowing both languages, than
memcpy(printline, employee_record, sizeof(employee_record));
printline[sizeof(employee_record)] = '\0';
They both describe what's occuring on an instruction level, but say nothing about what's happening on a purpose level. I've seen many Cobol programs comprised of a bunch of highly readable instructions, which when taken as a whole, seem meaningless.

Orthodox use of Hungarian merely guarantees that wrong types won't be assigned or passed, but does nothing toward revealing the underlying design.

This is the Summer of Code

Code seems to be making a comeback. Years of "non-coding drag and drop" development environments have done little for productivity, but cost a fortune in software and training. Java, Perl, Python, TCL/TK and GNU C and C++ are increasingly building the infrastructure of the 'net, and it's only a matter of time before "the enterprise" gets the message. And when they do, self-documenting code will be a powerful portfolio piece. And of course, you can make your drag and drop "code" self-documenting too.

Gracie Slick is an old lady, Culture Club is a speck on music history's dustheap, and text format source code is now catching its second wind.

By Steve Litt
You've noticed the June and July Troubleshooting Professional Magazines have been split into separate single-article .htm files, complete with next and previous links at the top and bottom of the article. I did that to conserve bandwidth, on the theory that most visitors are not reading every article.

I still author the mag as a single web page. But when I'm all done, I can run Perl program on the single file to create the individual ones, create the table of contents, and create all the forward and back links. I simply need to make sure that every article starts with an <H1> style title preceded by an anchor beginning with an underscore. The script takes care of everything else.

I wipped out this script in about 12 hours. I never intended it to be a showpiece. But I think it's a (small sized) example of the self-documentation methodologies I've previously discussed:
#!/usr/bin/perl -w
# by Steve Litt. Public domain. First published 7/26/1999.

use strict;

sub read_source_file
  my($fname) = "";
  $fname = $_[0];

  open(MYINPUTFILE, "<" . $fname) or die "Could not open " . $fname;
  my(@lines) = <MYINPUTFILE>;
sub make_contents_string
  my(@tags) = @{$_[0]};
  my(@titles) = @{$_[1]};
  my($contents) = "<center><b><font size=+3>CONTENTS</font></b></center><p><b><ul>\n";

  foreach $tag (@tags)
    $title = shift(@titles);
    if($tag =~ m/^_/)
      $contents = $contents . "<li><a href=\"$tag.htm\">" . $title ."</li>\n";
  $contents = $contents . "</ul></b><p>\n";  

sub get_issue_title
  my($page) = $_[0];
  $page =~ m/<title>(.*?)<\/title>/si;
  my($rtrn) = $1;
sub fill_article_lists
  my($oldpage) = $_[0];
  my(@tags, @titles, @starts, @lengths);  # LOCAL VERSIONS OF PASSED BACK ARGS
  my($prevStart) = 0;
  my($ss) = 0;

  push(@tags, "%startdoc");
  push(@titles, "STARTDOC");
  push(@starts, 0);
  while($oldpage =~ m/(<h1>[\n\w]*?<a NAME=")([_%].+?)("><\/a>)(.+?)(<\/h1>)/sg)
    if(!defined($1)) {last;}
    $start = pos($oldpage) - length($1) - length($2) - length($3) - length($4) - length($5);
    push(@tags, $2);
    push(@titles, $4);
    push(@starts, $start);
    push(@lengths, $start - $prevStart);
    $prevStart = $start;

  $oldpage =~ m/<\/body>/sig;
  $start = pos($oldpage);
  push(@lengths, $start - $prevStart);

  push(@tags, "%enddoc");
  push(@titles, "ENDDOC");
  push(@starts, $start);
  push(@lengths, length($oldpage) + 1 - $start);

  @{$_[1]} = @tags;
  @{$_[2]} = @titles;
  @{$_[3]} = @starts;
  @{$_[4]} = @lengths;

sub hdr_string
  my($title, $issuetitle) = @_;
  "<!doctype html public \"-//w3c//dtd html 4.0 transitional//en\">\n" .
  "<html>\n" .
  "<head>\n" .
  "   <meta http-equiv=\"Content-Type\" content=\"text/html; charset=iso-8859-1\">\n" .
  "   <title>$title</title>\n" .
  "</head>\n" .
  "<body text=\"#000000\" bgcolor=\"#FFFFFF\" link=\"#0000EE\" vlink=\"#551A8B\" alink=\"#FF0000\">\n" .
  "\n" .
  "<center><font size=+2><b><a href=\"../../troubleshooters.htm\">Troubleshooters.Com</a> Presents</b></font></center><p>\n" .
  "<center><font size=+1><b><a href=\"index.htm\">$issuetitle</a></b></font>\n" .
  "<b><font size=-2></font></b>\n" .
  "<p><b><font size=-2>Copyright (C) 1999 by Steve Litt. All rights reserved.\n" .
  "Materials from guest authors copyrighted by them and licensed for perpetual\n" .
  "use to Troubleshooting Professional Magazine. All rights reserved to the\n" .
  "copyright holder, except for items specifically marked otherwise (certain\n" .
  "free software source code, GNU/GPL, etc.). All material herein provided\n" .
  "\"As-Is\". User assumes all risk and responsibility for any outcome.</font></b></center>\n" .
sub write_article
  my($article)       = $_[0];
  my($ss)            = $_[1];
  my($issueTitle)    = $_[2];
  my($tag)           = $_[3];
  my(@tags)          = @{$_[4]};
  my(@titles)        = @{$_[5]};

  open(OUF, ">" . $tag . ".htm") or die "Can not write file for $tag";
  print OUF &hdr_string($titles[$ss], $issueTitle);

  print OUF "<center><table BORDER COLS=1 WIDTH=\"100%\" BGCOLOR=\"#FFCCCC\" ><tr><td><center><b><font size=-1>\n";
  if($ss <= 0)
    print OUF "<--";
  elsif(!($tags[$ss-1] =~ m/^_/))
    print OUF "<--";
    print OUF "<a href=\"$tags[$ss-1].htm\"><--$titles[$ss-1]</a>";
  print OUF "&nbsp;&nbsp;&nbsp;|&nbsp;&nbsp;";
  print OUF "<a href=\"./index.htm\">Contents</a>";
  print OUF "&nbsp;&nbsp;&nbsp;|&nbsp;&nbsp;&nbsp;";  
  if($ss >= scalar(@tags))
    print OUF "-->";
  elsif(!($tags[$ss+1] =~ m/^_/))
    print OUF "-->";
    print OUF "<a href=\"$tags[$ss+1].htm\">$titles[$ss+1]--></a>";
  print OUF "</font></b></center></td></tr></table></center>\n";
  print OUF "$article\n";
  print OUF "<hr WIDTH=\"100%\"><b><ul>\n";

  if($ss >= scalar(@tags))
    print OUF "<li>Next article:</li>\n";
  elsif(!($tags[$ss+1] =~ m/^_/))
    print OUF "<li>Next article:</li>\n";
    print OUF "<li><a href=\"$tags[$ss+1].htm\">Next article: $titles[$ss+1]</a></li>\n";
  if($ss <= 0)
    print OUF "<li>Previous article:</li>\n";
  elsif(!($tags[$ss-1] =~ m/^_/))
    print OUF "<li>Previous article:</li>\n";
    print OUF "<li><a href=\"$tags[$ss-1].htm\">Previous article: $titles[$ss-1]</a></li>\n";
  print OUF "<li><a href=\"index.htm\">Magazine Contents</a></li>\n";
  print OUF "<li><font size=+1><a href=\"../../troubleshooters.htm\">Troubleshooters.Com</a></font></li>\n";
  print OUF "</ul></b>";

  print OUF "<\/body><\/html>\n";

sub writem
  my($page)       =   $_[0];
  my($infname)    =   $_[1];
  my($issueTitle) =   $_[2];
  my(@tags)       = @{$_[3]};
  my(@titles)     = @{$_[4]};
  my(@starts)     = @{$_[5]};
  my(@lengths)    = @{$_[6]};

  my($ss) = 0;

  open(MAIN, ">index.htm") or die "Cannot open index.htm";

  foreach $tag (@tags)
    my($article) = substr($page,$starts[$ss],$lengths[$ss]);
    if($tag =~ m/^_/)
    elsif($tag eq "%startdoc")
      print MAIN $article;
    elsif($tag eq "%contents")
      print MAIN &make_contents_string(\@tags, \@titles);
    elsif($tag eq "%enddoc")
      print MAIN "</body></html>\n";
      die "Cannot print undefined tag.";
    $ss = $ss + 1;

sub main()
  my($srcfile) = $ARGV[0];
  my(@tags, @titles, @starts, @lengths);
  my(@sourcelines) = &read_source_file($srcfile);
  if(!defined(@sourcelines)){die "Must have source file as single argument";}
  my($sourcepage) = join("", @sourcelines);

  &fill_article_lists($sourcepage, \@tags, \@titles, \@starts, \@lengths);

  my($issueTitle) = &get_issue_title($sourcepage);


Let's examine The first and fourth lines of main() document the command syntax. Basically the main routine says read the source file and mold into a string called $sourcepage. Fill the four article lists, get the issue title, then write the article pages. Subroutine writem() should have been called write_web_pages(). All in all a fine, self documenting main routine.

Routines read_source_file(), get_issue_title() and fill_article_lists() are pretty obvious (though the latter needs a few comments for complete readability). Poorly named writem() starts out readable, identifying its arguments and an output file, but gets a little hairy in the foreach $tag (@tags) loop. But it's not too bad, and at least identifies @tags as the controlling entity. writem() calls write_article() with six arguments.

All pretense of readability is dropped in write_article(). It does an admirable job identifying the arguments and opening the output file, but then descends into a conglomeration of code, which comments tell us perform three sequential tasks: 1)Create page top pointers, 2)Create the page itself, 3)Create page bottom pointers. Note that #2 is trivial. What I should have done is had #1 and #3 be subroutine calls print_page_top() and print_page_bottom() respectively. The problem is I didn't want to pass all those darn arguments again, and all those lists and subscripts were getting confusing.

In retrospect, of course, the solution was to encapsulate lists @tags, @titles, @starts, @lengths and subscript $ss into an object with a name like $articles. $articles might have methods like next(), previous(), tag(), title(), start(), length(), getSubscript(), setSubscript(). The constructor would take the main routine's $sourcepage as an argument and basically do the same parsing as the existing fill_article_lists(). Such a construction would have cut arguments passed to routines, and encouraged further functional decomposition, resulting in vastly improved readability approaching true self-documentation.

Steve Litt can be reached at Steve Litt's email address.

A Better Blacksmith Tool?

By Steve Litt
Aren't these tips a little late? After all, the days of monolithic blocks of code are gone. Now we drag, drop, and then code events. Right?

Everybody knows all real app dev is done with Delphi, Clarion, VB, Powerbuilder, JBuilder, Visual J++, PowerJ, Visual Fox and the like. These days we must have rapid development. Only the backward looking dinosaurs use text source code. These tips would have been useful in 1990, but they're a day late and a dollar short.

Well, except that they can be used in drag n drop RAD. Most of these RAD environments force you to design your program objects around screens. Ugh. (And thank you Powerbuilder and Clarion for not being like that). But even so, considering that every screen *should* have some function, correct naming of the screens and limitations of functionality will produce (drum roll please) self documenting code. And even in the most RADical RAD, you can design a complex object around a "hello world" style screen. So if you're a good designer, the tips outlined earlier in this Troubleshooting Professional Magazine issue produce self documenting RAD.

And then there's the fact that the fundamental premise of this article might be false. Is source code dead? Was Linux written in VB? Or even VC++ with MFC? Or even C++? Nope, C. C for the kernal, C for enhancements, C for drivers. But that's all systems programming. You can't do apps in code, right?

Try tcl/tk -- compare its productivity to VB. Don't like tcl's unusual syntax? Program in Python or Perl, and use their tk modules for screens. RAD code is not an oxymoron.

But certainly as Linux matures drag and drop environments will displace code. Right? Well, that's really two questions. Will drag and drop environments appear? Certainly. With free Python, Perl and GNU C, the d&d environment is a proprietorists last bastion of price gouging. Will d&d displace code? The Linux marketplace has a proven track record of knowing when the king has no clothes. Only time will tell.

And all of this may be a moot argument. Because Linux, with its no-hassle licensing, modularity, and wealth of utilities, just may tip the scales toward the software-manufacturing model the deified but not achieved the last fifteen years. Read on...

Steve Litt can be reached at Steve Litt's email address.

Linux Log: The Lesson of the Artist

Linux Log is a regular column in Troubleshooting Professional Magazine, authored by Steve Litt. Each month we'll explore a facet of Linux as it relates to that month's theme.
$80.00/hr isn't a bad hourly, is it? Not the best, but not bad. Now turn the clock back to 1994, with recession still strangling California. $80/hr starts sounding much better. Imagine a *street artist* making $80/hr in 1994 Port Hueneme, California. It happened. I saw it. And I vowed to learn from it.

His Latin Lover looks showed through the gas mask worn as protection from paint fumes. He knew maybe 200 words of English. Working with spray paint, stencils and rags on the sidewalk of Port Hueneme's tourist area, he painted the most beautiful ocean and sky scenes imaginable. Every one custom made for the customer. In 15 minutes. For $20.00. I saw it, I timed it. Every 15 minutes, another $20.00 check. $80.00/hr.

The women in the 100+ person crowd surrounding The Artist raved about the beautiful paintings, and how much they wanted one. The men, every one of whom had done the same timings and calculations as I, wondered aloud if they had chosen the right career. Being that wild and crazy process-oriented animal I am, I vowed to generalize his techniques to help me with my programming and tech-writing.

I watched him for hours. Bought one of his paintings. Dropped all other thoughts, and pondered what I'd seen. By nightfall I'd synthasized his production methodology down to goal, tools and riffs. Within a month I was using goal, tools and riffs to double my tech-writing productivity. Programming took much longer, and I'll talk more on that later in this article. But right now, let's talk about goal, tools and riffs.


He'd spend a minute interviewing the customer as to what she (most but not all were female) wanted. There would be a little back and forth as he molded her vision into what his tools and riffs allowed. Everyone got what they perceived as a custom painting, but all the paintings were actually within some pretty tight parameters. No skyscrapers, no football players. Just ocean, sky, fish, birds, sun, stars, air and water. In the first minute or two, he knew exactly what was to be painted, and had placed it in the framework of what his tools and riffs would allow. He had a goal.


Watching the man work with his tools was shear joy. Stencils for every kind of bird and fish. Stencils for entire flocks of birds and schools of fish. Various paint brushes and cans of spray paint. Rags. His tools were his life. He knew them like a mother knows her child. He knew where each was. After each use it was appropriately cleaned and put back in exactly the right place. He never needed to search for a tool -- he just grabbed, used, cleaned and put back.


It's a poor workman who blames his tools. And a poor workman with superior tools does poor work. Tools are nothing unless used well. And who better to elevate the use of tools to an artform than The Artist? He accessed his tools with riffs.

Riff is a musical term for a sequence of notes so familiar to the musician's fingers as to be a reflex. That's part of the reason musicians play scales -- riffs come from scales. Go into your old record collection (or your parents or grandparents) and pull out some Led Zeppelin records. Zep took a riff, beat it to death, and made a killer song. Listen and see if you don't agree. If you have no Led Zeppelin, listen to any Heavy Metal. Most Metal is riff based to a degree.

Not musically inclined? No problem. Do you touch type? If so, I bet you don't type the word "the" as three letters, do you? It's a quick reflex finger movement. A riff.

The Artists riffs were richer and used varying tools. Need a school of multi-colored fish? Whip out a fish school stencil and can of spray paint to make the bodies. Center another stencil and grab a different color spray paint to paint the fishes highlights. Elapsed time -- 10 seconds.

Need the sun? Spray a large blob of yellow. Place an empty spray can upright on the yellow, and spray sky color all around. Lift the can, and smear with a rag to get the sun's radience. Reflex. Elapsed time -- 20 seconds.

Need a comet? Spray white into a round stencil, lift and smear upwards with a rag. Elapsed time -- 5 seconds.

He's used these riffs thousands of times. From these limited stock riffs he creates thousands of seemingly custom paintings.

My Application Development Failure

As I mentioned, within a month I'd doubled tech writing productivity using The Artists methodology. Goal, tools and riffs have transformed authoring and marketing of Troubleshooters.Com into a simple writing task. But developing apps with The Artists methodologies continues to elude me. Why?

The Artist created individual paintings. He didn't have to match an existing decor. The customer didn't tell him what tools he had to use. The customer didn't make The Artist paint in a way that could be easily modified. The Artist had freedoms the application developer just doesn't have.

Let's face it. Very few of us automate a business from scratch. They have a DBMS, so we have to hit that DBMS. They have an OS, so we have to be compatible with that OS. They have an existing software inventory written in language X, and programmers familiar with language X, so we have to write in language X. Of course using framework Y. Tens of languages, each with tens of frameworks. As we move from job to job, task to task, contract to contract, our knowledge of tools becomes broad, but seldom deep.

Linux to the Rescue

GPL (GNU General Public License, "free software") changes everything. You can "own" hundreds of free, reusable, tested tools. Without the motivation to "differentiate" brought on by the intellectual property sales model, everyone pretty much contributes to the best tool of each genre. The tools easily interconnect, often via pipes, backticks and command line args. An awk/sed expert can replace a hundred lines of C with a five line shelscript. Linux has commands for data compression, image conversion, text handling, paper formatting, and almost anything you can think of. Connect them together using Python.

And speaking of Python, it comes with modules for html, xml, ftp, math, strings, database, complex numbers for you engineering types, system calls. And of course the Tkinter interface to tcl/tk, so you can build nice little gui apps with Python. Web app? No problem -- use Python-centric Zope or quick and simple PHP. The technologist fully conversant with these tools can punch out a sophisticated custom app in a day. An app built from pre-built, intensively tested, fully modularized (unlike most "guis" and "frameworks) components.

Perhaps most promising, open source vertical market components are appearing. There is now an open source project to develop medical practice management software. Can litigation support, retail and hospitality software be far behind?

So What?

There's no market for sophisticated one day apps in this market-busting summer of 1999. With money to burn, software shops and IT departments are going full MS Windows with $1500.00 per seat GUI development environments. And with this summer's programmer (scuse me, developer) shortage, anyone capable of writing a one day app is working for the big spenders. And many believe this will last forever.

He who forgets history is bound to repeat it. There have been more than ten complete economic cycles this century, and nothing's happened to change that trend. But the number of developers has doubled the past few years. And when the downturn comes, the newly minted VB developers, Java jockeys, Visual C++ gurus, Crystal Report savants and Y2K mechanics won't simply walk into the sunset. They'll compete. A brutal game of musical chairs, with large numbers of programmers indefinitely kicked out of the glass house.

Most dispossessed developers will become "starving artists". But a few will follow the footsteps of that long ago street artist. They'll look at the startup small businesses, so long denied automation due to bid-up developer costs, as a new market. Even in a recession, a healthy small business can afford $640 for a custom app. And since the programmer builds it from tried and true open source components, it takes a day. That's $80/hr, in a recession, with their less enterprising brethren on the unemployment rolls.

But it gets better. "The Enterprise", no longer able to afford the Microsoft tax or the $1500/seat development app costs, or the training costs of ever changing and buggy development environments, will take note of the small business successes. The goal/tools/riffs open source pioneers will be invited into the department, then the enterprise. And that's how Linux will finally win.

Wishful Thinking

Yeah, sure. The "enterprise" will dump their costly inventory of Microsoft centric software for Linux centric. It will never happen. Will it?

The 1980's PC revolution is an exact precident. Sick to death of mainframe and mini vendors and their own glass house holding them over a barrel, departments hired PC pioneers to quickly implement software solutions for the department. As soon as the glass house gang had superior development competition, self preservation dictated they hire PC programmers. A skeptic might say the 80's revolution was about a new hardware platform. Not really. It was about price, reliability, development speed, training cost, and about a widespread uniform easily programmable operating system.

On Your Mark

It's gonna happen. Open source software is rapidly catching up to, and even overtaking, proprietary. Linux is squeezing out NT at the department/SOHO/small business server level, and it's only a matter of time before that spreads both directions to the enterprise and the desktop. And our present boom economy? History shows that's just about over, and with the Y2K spectre close on the horizon, it's never too soon to grab those tools and start practicing those riffs. Getting there early differentiates those who follow The Artist from those who become starving artists.
Steve Litt can be reached at Steve Litt's email address.

Letters to the Editor

All letters become the property of the publisher (Steve Litt), and may be edited for clarity or brevity. We especially welcome additions, clarifications, corrections or flames from vendors whose products have been reviewed in this magazine. We reserve the right to not publish letters we deem in bad taste (bad language, obscenity, hate, lewd, violence, etc.).
Submit letters to the editor to Steve Litt's email address, and be sure the subject reads "Letter to the Editor". We regret that we cannot return your letter, so please make a copy of it for future reference.

How to Submit an Article

We anticipate two to five articles per issue, with issues coming out monthly. We look for articles that pertain to the Troubleshooting Process. This can be done as an essay, with humor, with a case study, or some other literary device. A Troubleshooting poem would be nice. Submissions may mention a specific product, but must be useful without the purchase of that product. Content must greatly overpower advertising. Submissions should be between 250 and 2000 words long.

All submissions become the property of the publisher (Steve Litt), unless other arrangements are previously made in writing. We do not currently pay for articles. Troubleshooters.Com reserves the right to edit any submission for clarity or brevity. Any published article will include a two sentence description of the author, a hypertext link to his or her email, and a phone number if desired. Upon request, we will include a hypertext link, at the end of the magazine issue, to the author's website, providing that website meets the Troubleshooters.Com criteria for links and that the author's website first links to Troubleshooters.Com.

Submissions should be emailed to Steve Litt's email address, with subject line Article Submission. The first paragraph of your message should read as follows (unless other arrangements are previously made in writing):

I (your name), am submitting this article for possible publication in Troubleshooters.Com. I understand that this submission becomes the property of the publisher, Steve Litt, whether or not it is published, and that Steve Litt reserves the right to edit my submission for clarity or brevity. I certify that I wrote this submission and no part of it is owned by, written by or copyrighted by others.
After that paragraph, write the title, text of the article, and a two sentence description of the author.

URLs Mentioned in this Issue