Speed Ream's Slashdot Article

The following is a reformatted version of my brother's article about Leo that appeared at slashdot.  The original version of that article is here.  This is one of the best descriptions of why Leo is special:

Hello, my full name is David Speed Ream. I am known as Speed to friends and enemies alike, but I gladly answer to David or most any other handle. I am an unabashed and biased fan of Leo, the fact that it was written by my brother Edward only slightly coloring my already colored glasses. I have been testing and using Leo in software production for over 4 years. My company currently has over 50,000 lines of code in over 100 source files that are written using Leo.

My comments are from two points of view, the first being software project manager for a complicated, multi-module software product, and the second being as a production line coder. For me, Leo’s greatest and only real drawback is the learning curve. This learning curve can be shallow is if all that is required is that someone code using Leo. However, in our company we allocate 40 to 80 hours on top of the normal coding load for someone to come up to speed on Leo. The ROI (return on investment) is calculated by me to be on the order of 3 months. So if I hire a consultant for less than 3 months, I don’t teach him Leo, even though all source code in our company must reside in Leo files for the reasons I won’t go into now.

I consider that my coders are 15 to 30 percent more efficient in their daily operations than my competition’s people. This indefensible claim of mine is based on the changes in my productivity as Leo grew from a test document production tool to the primary production method for all our assembly, c and cpp source code.

Personally, I hate to deal with documentation when I write code, except: 1) When I am first sitting down to solve a new problem. Then the documentation becomes quite long-winded and pontificatory, as if I were the only one on earth smart enough to solve the problem - or 2) When I come back to code I or someone else has written and find the documentation insufficient to understand the code without study (seems to be most of the time).

So I do not require my engineers or myself to do a great job of documentation, nor do I use Leo for that purpose. Rather, it is Leo’s outlining and organizing ability, and Leo’s ability to create source files from within the outline that give me what I think is a tremendous competitive advantage. Each of my company’s products run on all versions of windows from Win 3.1 to XP. In our flagship software piece, there are ten main modules, and each module is maintained by one single Leo file. In the codec module, one Leo file named compress.leo organizes and creates seven .asm files, forty-four .c files, twenty .h files, two .def files, four .mak files, etc. etc. etc. This one file can be checked out from source code control and given to an engineer for the addition of a new feature.

In it are contained all the known issues for the code, each issue arranged in its own clone section. One clone section groups together every routine, variable or type definition that must change between different versions of Windows. These sections could be from six different c source files, two assembly files, and eight .h files. Another clone section groups together those sections relating to memory problems, which change according to the memory configuration and TSR configuration (or lack thereof) on the target machine. Another clone section groups sections that fail (or don’t fail) if the routine in question was accidentally run during the dreaded ‘interrupt time’. Another clone section is a section containing clones, each of which is named after the major bug that was fixed when the engineer who fixed the bug grouped a bunch of routines, definitions, etc. together to fix the bug.

None of the above clone sections was ‘designed’ into the document. Just the opposite happens. When the codec was first written, there was just a single Leo file with a bunch of sections for each c routine or assembly module. As the product grew and was tested on various platforms, each failure of the module was organized into clones each time a failure was fixed. This is what I call self documenting code. This has nothing to do with me sitting and documenting anything. Its just that the structure of a bug fix (or product enhancement) lives on long after the coding is done, as long as no one is foolish enough to delete the cloned sections that document what happened.

In actual practice, this organizational ‘history’ is so powerful that I can’t begin to describe it. A reverse learning curve happens when an engineer gets a Leo file that already has the ‘interrupt time sensitive’ routines grouped together by the last unfortunate soul who had to work on them. There may not be any more written documentation, but the knowledge contained in the structure can be breathtaking. It is certainly timesaving. I find this particularly true in my own case. Often I’ll look at some code that seems totally unfamiliar and think ‘what idiot wrote this crap’. Then I’ll look at the version control comments and realize that I wrote the crap. Then for sure I know the documentation is non-existent, but the clones I used to develop it are still there, and they always serve to refresh my memory in an indescribable way.

Enough of this commentary, I just looked at the clock. Best wishes to anyone willing to try Leo for a week. I hope you will be glad you did.