Some walls are necessary. We use brick-and-mortar walls to support buildings and firewalls to protect our computers from attack. But not all walls are good. Consider the Berlin Wall, a wall of segregation. It divided a country and its citizens, but has subsequently been brought down by people working together because upon re-evaluation the Wall did more harm than good. These thoughts led me to think about the walls that developers, QA managers, and database professionals have erected over the years to segregate themselves. Is it time to re-evaluate our own walls?

Most organizations, from the small mom-and-pops to the large enterprises, have established all kinds of walls. An example of a good wall is one that is intended to stop or prevent unwarranted access to mission-critical systems. An example of a bad wall would be one that stifles communication or understanding of the data being manipulated. Originally all walls were put in place to save time and rework, but over time it's a good idea to review long-held practices and reassess what's still effective and what isn't.

I've had a varied career; I have been a lone developer, a member of a team, a project manager, and I've run R&D for organizations. It was always been a practice to keep the database in the database professional's hands and away from others, such as application developers and QA staff. Why did I (and as far as I know, the rest of the IT industry) do this?

The answer is simple: the database is always a key fixture in any organization. Since the beginning, companies understood that they needed to have superior applications to collect information; the result was an explosion of invaluable data and information. And the industry learned quickly how bad things could get if that data was corrupted or lost. Based on this fear we established artificial walls that would isolate access, establish access rights, and limit communications between various departments in an organization.

But that was then. Is the same fear-based paradigm still valid in today's IT environment?

Various studies over the years have found that developers with correct requirements and the right tools are more productive and deliver better solutions. This makes sense, and it's been proven to be true across professions such as construction or automobile repair. Over the years, one of the areas that has been lacking in development are tools that cross boundaries to help developers write better SQL and database-related code. In many instances, the initial database interaction created by the developer is little more than a placeholder for the DBA or database professional to review later. This approach is inherently flawed primarily because it enables early mistakes to make it through the entire process to production.

We have all entered mistyped data, which has a cascading effect on the capture and storage of information. Developers who do not understand the database structures will inadvertently create new tables that are sometimes redundant, and don't always write the most effective SQL code. Somewhere in the chain the review process breaks, and mistakes that shouldn't happen do. This leads to DBAs missing or breaking SLA(s), mistrust between developers and DBAs and the application not working as it should.

So doesn't it make sense to put things in place that, at a minimum, help ensure the application developer doesn't just put in some "holder" code, but instead has a fairly good chance of creating a database manipulation layer that if the review breaks down, the code created will work as expected and that most errors or inefficiencies are caught and fixed upfront?

Current Practices: Barricades to Effectiveness
Developers don't want to write code that's incorrect, not up to the best standards, or performs poorly. So having the tools to help developers write solid code is essential. There are next to no developers in the world who think debuggers are bad or believe they shouldn't be used to isolate problems in code. Some problems are easily found using a debugger rather than applying other bug-hunting techniques. The same is true when it comes to optimizing an application.

A long-standing debate exists about the best approach to optimize for speed. Some in the developer community believe that optimization should only happen after the initial QA process has certified the application. Others have learned that making changes for speed and efficiency that late in the process can be dangerous and that it often takes another full QA cycle to ensure that the optimization process didn't break anything. (All developers have been bitten by the "one-line" change that fixes a problem but causes 50 other issues.)

Others in the community believe that initial testing of the software should be optimized before reaching QA. This can be effective, but also means code needs be reworked since QA hasn't certified it as correct. Then there's a school that believes optimization should only happen when the application is in maintenance, since that's the only time the software is usually considered correct.

Most likely an "all of these" approach is warranted. Consider Wall Street: if it didn't build performance and QA into every step its ability to process huge numbers of transactions would be almost impossible, and disastrous code could make it into production. Optimization needs to occur when things can be isolated, so using profiling tools throughout the development lifecycle is important for the developer. For example, how would a developer test the performance of an ‘x' algorithm versus a ‘y' algorithm if he doesn't have the tools to profile it in the first place?

It's been an established best practice that finding bugs early in the development process is a good thing. It saves time, and eliminates rework, retest, and redeploy cycles. It also reduces the possibility of bad code, embarrassing the company and delaying delivery.

Standardized Tooling: Building Bridges, Not Walls
Giving developers advanced tools to help write cleaner, better-performing code is essential in today's fast-paced marketplace. Plus, many organizations have put extra pressure on developers and DBAs, combining roles and making them wear both hats. But, there's a wall we've erected that's causing problems; the tools are not consistent between the two worlds.

But there are new tools on the market that help developers and DBAs visualize rather than read what they are building, and more importantly, reverse-engineer what they've already built. Products that let complex concepts be modeled and discussed from both an application developer, Q/A, DBA, and data center perspective have become critical to organizations. Existing tools will help find nightmares in applications, but the new tools are adding suggestions, best practices, and help in fixing code on the spot, even considering the context of who is making the change. Insights that may not be useful or understood from a developer standpoint may be incredibly important to a DBA and thus solve problems in less time. These new tools have capabilities we only dreamed about a few years ago, such as profiling the entire lifecycle so no one area is left out of the information chain, as well as automatically generating documentation, portals, and Wiki support to expose information to key players.

The walls that have been put up over the years are at least getting fitted with a few doors and windows; and in some cases they are even being taken down brick by brick, allowing for clean and nimble development. Will these changes eliminate all the problems facing developers and database professionals? Of course not, but hopefully standardized tooling will, at a minimum, increase the "vernacular" and reduce the bad code making it to production.