Rebuilding the Enterprise - Software, Hardware and Peopleware Migrations for the Systems Architect

6/29/2014

Technical Debt & the Lazy Developer

6/29/2014 Posted by William Berry , No comments
I have been reflecting over the last few days, what it means to be a "lazy" developer, inclusive of both positive and negative connotations.

Yup, that was me, and what I see is manifest in two distinct ways:
  1. Lazy Loading
  2. Misaligned priorities, unwillingness to contribute and apathy
Each or these points is directly resulting in the accumulation of technical debt.  These thoughts are by no means localized to software development, the industry just happens to be an easy vehicle of context.

***

From my vantage point, lazy loading is that delightful form of efficient procrastination where I don't actually do something until it's needed.  I don't add interfaces until the testing requires it.  I don't leverage encapsulation until I am passing around List<Tuple<int, KeyValuePair<string, string>>>.  If the code can be simple, flat, clean and free of excess abstractions ... then I leave it that way.

There is also the classic example of lazy loading ... ORMs; the perfect double edged sword.  The vast majority of the time, lazy loading works to your benefit; but, every now and again ... it totally screws you.

Where I find myself most guilty of lazy development is the classic arenas: instrumentation, telemetry data collection, error handling, logging, performance optimization and documentation.

The first four, instrumentation, telemetry data collection, error handling and logging, are all classic after thoughts.  Even the most well intentioned developer will persistently screw this up.  Why? Because you don't NEED these things to get a product up and functioning; except that maturity and experience will eventually teach you, through enough browbeating, that you do in fact NEED these things to get a product up and functioning.

On the surface, it's clear how one can address this form of technical debt.  You "should" have logging, you "should" be collecting telemetry data, and at some point you will address the debt and add the elements, simply because you are left with no other choice.  But, your debts from implementation after fact, will not net 0.  You will be left with a very subtle issue.  That, I posit, can cause nearly as much damage to a system/product.

When rings like logging and error handling are addressed after the core of the application has been solidified, they are merely stapled on features.  They may have been woven into the fabric of the system, but they have not grown and matured with the core entities.  As such, they will never feel like a natural part of the development process and there will always be friction in the usage of them.

I have experienced refactoring several existing projects, as well as building a few new ones, where I swore I would not lazy-load the instrumentation ... surprise, still haven't learned (and I bet I am not the only with this issue).

On the flip side, by most accounts, early performance optimization is a waste of time.  Don't fix it, if it ain't broke, right?  What I don't think has been settled officially is when "early", is.  I suggest, that performance go hand in hand with instrumentation and collection of telemetry data.  By leaving performance optimization and the supporting framework for a time when you actually have the problems, you're open to the exact same issue noted above, only subtly worse.

Not only will you be gluing your performance data collection to the side of your app, rather than making it integral, you have no historical data to help you key in on when things went gone wrong.  You could have had video of all the events leading up to and beyond the tragedy, instead, you are left in a room with no windows, no doors and no murder weapon.  While I will concede that early performance optimization can be a wasted effort, the establishment of a supporting architecture and integration of carefully selected tooling, is by no means a waste of time.

There are two main things in flux here, with the first being tooling.  Adding instrumentation, data collection, logging, error handling and the like, is not easy; and, there are typically several non-trivial architectural decisions that must be made upfront for their inclusion to feel native, and therefore inspire usage.  Compound the complexity of tooling integrations with the maturity required to identify the situations upfront ... it's no wonder why we are lazy with the implementation of these components and why so many projects suffer from lazy loaded technical debt.

***


Most people easily envision the technical debt that is the byproduct of "common" laziness: poor design choices, deliberately evading standard approaches, partially implemented or hack solutions, etc.  These manifestations are actually pretty easy to clean up.

If a piece of code is poorly factored or does not adhere to typical implementation strategies, then simply mix some refactoring with training and education, and more often than not, the problem goes away with little fanfare.

Another easily identified negative manifestation of laziness is weak or missing documentation.  While its quite obvious that most all projects need documentation to capture both institutional knowledge and execution context, what is not always obvious is the emergence of technical debt outside of the software product.

Poor documentation will directly impact marketing efforts and project reputation.  Unless the project is in it's infancy, this debt is not easily addressable by simple bolstering the amount/quality of the documentation.  Often, trust of the end user or implementing developer has been lost or compromised, and interest payments here can be very costly.

One technique for small teams to address documentation issues up front, is to communicate through formal specifications.  Though the Agile Manifesto favors working software over comprehensive documentation, what we are really trying to capture here is execution context, design decisions, encapsulation of complexity, domain modeling and the development of a ubiquitous language.

By codifying in documentation the initial project activities, you are setting a platform upon which to build later.  As the project begins to mature and ramp up, leave the hard documentation behind in favor of patterns and practices that will result in automatic document generation as the project moves forward.

To give a hard example, let's say we were building a internal facing API over a data store.  We would start with a simple requirements or story document that laid out the 10,000 foot overview.  The document would include some goals for performance, usage patterns, etc.  Next we would generate the core resource endpoints and entity models, thereby distilling entity properties and furthering the development of our ubiquitous language.  Once the core resources have been loosely agreed up, the team simply begins development.

That said, one of the first features/stories to be implemented is a management interface that auto-documents the API's entities and resource URLs.  Using this approach we have leveraged documentation to record foundational execution context and provided a mechanism to ensure the consumer's documentation is always up to date and accurate.  Additionally, we have mitigating the technical debt thrown off by weak documentation and through automation we have ensured that the product and its documentation can grow and mature in unison.

The devil however, in all this laziness, is as manifest from apathy or misaligned values.  In both cases the technical debt is accruing in not just the team's current work product, but also in everything the team touches.

Of the two, apathy is significantly easier to mitigate.  Though it can spread like a virus, the "I don't give a shit" attitude, 9 times out of 10, is easily course corrected when caught early or when isolated to a single team member.  The technical debt accrual in the products, much like in the simple laziness example above, can almost alway be brought back into control, through an ernest refactoring effort.

The complexity in dealing with apathy is addressing the team's underlaying issues.  Understanding the root cause of the behavioral shift is paramount.  One technique that a manager can leverage is to actually join the tumult.  Through participation, the instigator(s) and core issue(s) is often exposed.

Having been both an instigator and a manager dealing with rampant apathy, I was often able to attribute it to long work hours, and poor training/tooling.  The savvy team lead will leverage the apathy to unite the team around the core issue and then behind the scenes facilitate its resolution ... building team cohesion and course correcting all in one move.

The final point of contention is to address the technical debt thrown off by laziness as it is manifest through misaligned values.  Though intention and commitment may be present, there may be an unwillingness to work with the values or towards the goals of the rest of the team; the result being a single product headed in multiple directions.

The laziness associated with misaligned values is not the fault of the seemingly misdirected employee, but rather with the project lead or the instigator of change.  As is well documented in Leading Change by Kotter - the project lead or change initiator is at fault here, simply because they have failed to instill a sufficient sense of urgency to inspire change.  Without reason or sufficient energy, all systems will remain in stasis.

If misaligned values or divergent paths is at the core of your technical debt generation, then I highly recommend reading Leading Change.

The approach to addressing the technical debt from misaligned values is complex.  Not only will you be leading a change effort to adjust team alignment, but you will need to wait to drive the refactorings until after the team has unified it's values.  Any significant addressing of technical debt before alignment, will result in continued divergence, albeit with a smaller delta.

Being cognizant of laziness within your organization and/or teams is the first stepping stone.  Understanding it's various manifestations and coping with or mitigating the side effects is the trump card.  Lastly, don't forget to turn the mirror on yourself.  Are you leading change effectively, setting the guidelines and goals and innovating by example ... if not, don't try to fix your team, before you have taken the time to fix yourself. 

I would appreciate comments and a conversation either here on twitter where you can reach me @WilliamBerryiii

0 comments:

Post a Comment