George Mauer is on the Net

Oooh! I Wrote Something for Upper Management

Today I was asked to write up a short justification for a rewrite of a godawful piece of software that I have been complaining about to anyone who would listen.  Yay!  Inspired by essays I have been reading by Jim Shore I might have been a bit too agressive.  But here’s the final(?) version:

Proposal for a Rewrite of the {Program Name} Software
The purpose of software is to make our lives easier; it is a tool like a hammer or a screwdriver and far more malleable than either. This is generally a good thing as it allows us to mold it to our specific needs and particular processes. If software is a screwdriver meant to do screwdriver-type activities it should be with a minimal amount of effort that we can get it to work on a phillips or a flat-head screw. Unfortunately even screwdrivers can be poorly made, their handles can disintegrate as we apply too much torque, they can be made to poor standards so they don’t grip the screw quite right and slip frequently, and they can be made out of cheap material so that any attempts to remedy the above only compounds the problems as the shaft bends and warps in unexpected ways. Without taking the metaphor any further, it is my professional opinion that – external appearances aside – our current {Program Name} Software is this sort of unwieldy tool and in need of an urgent re-working.
Current Problems and Future Implications
The current implementation of our {Program Name} software has a multitude of problems. First and foremost, it seems to have been created as a prototype rather than a lasting project. As such, it is not meant to be extended to work in the ways that we need it to work. It is also not possible to use it with automated testing software to check the various contingencies we need it to work under and to ensure that future features do not break already existing ones. Without the ability to write automated self-tests, keeping the system operating across the conditions at the various terminals and feed company locations will continue to become more and more of a burden that will continue depleting our already low development time.
Furthermore, under the current system, when a bug is encountered, it is notoriously difficult to troubleshoot. For a problem at a site, our only recourse is to attempt to recreate it by simulating the conditions from user testimony. This is a painstaking and highly uncertain process. Take as example the multitudes of issues caused by poor network connectivity. The current system has no plan for dealing with short-term outages which, while inexistent when the software is developed in-house, are ever-present in the field. As such, it frequently ends up acting in bizarre ways that get reported infrequently and incompletely. Under these conditions they can be nearly impossible to repair. The software was simply not produced with the idea that it would require in-the-field maintenance and diagnosis; although expedient attempts to remedy this situation have been made, deep architectural flaws severely limit their possibilities.
The bottom line is that behind-the-scenes the {Program Name} is poorly designed and implemented. Much of the inner-workings are extremely confusing and no attempt has been made to consistently use established industry or even the developer’s own standards. There is no documentation anywhere. The problems are deep-rooted enough to need to be addressed immediately. In saying this I want to make it absolutely clear that in my opinion our difficulties arise not from the inescapable need to add features, but from the fact that the scope of the initial design completely omitted considerations of clarity, extensibility and in-the-field testing.
Proposed Changes
I propose a complete redesign of the {Program Name} system with an emphasis on the creation of a system that is simple, easy to maintain, extend, and troubleshoot. We will of course address all of our current needs but we will use established techniques to create software that works well with a model of continuous development. The new system would also be designed to streamline troubleshooting and include automatic self-testing to ensure that it is shipped out with as few bugs as possible. We will achieve this by sticking close to C# .NET best practices making the application truly object oriented, using Agile planning techniques and test-driven development, and optionally integrating with a dependency injection framework to ensure extensibility. While the changes experienced on the user side will be minimal, the payoff on the development and maintenance side will be almost immediate. I fully expect development time under a redesigned system to shrink by at least a factor of four, for new deployments to become seamless instead of taking days to configure properly, and for maintenance time to fall from over 5 hours per week to well under one. As a tentative schedule I would propose the following: Half a week of design (much of it has already been done), one to two weeks of implementation and a week of testing. I believe that by working on this project 80-90% of my time I should be able to finish it and have it ready to go in three to four weeks.

Thank You Very Much,
George Mauer
Software Developer


May 5, 2008 - Posted by | Programming | , ,

1 Comment »

  1. George,

    If this was for executives, they probably didn’t read it. Try reducing it to on eintroductory sentence to establish the context, five bullet points to list the main problems, and one summary sentence to tell them what you recommend and how much money it will save them if they agree to let you do it.


    Comment by Steven A. Lowe | September 30, 2008 | Reply

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google+ photo

You are commenting using your Google+ account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )


Connecting to %s

%d bloggers like this: