Apply today for a FREE subscription to CIO Magazine!
Sun, Jun 17, 2007 15:58 EDT
|
Posted by: Michael Kavis in Soapbox Topic: DevelopmentBlog: Delivering the Goods
Current Rating: |
Maybe I am too old school, but I am seeing a disturbing trend in software development these days.
I see development efforts that focus almost entirely on business requirements with disregard to technical requirements and real software design. Back in the day, computer resources were limited and you had to design applications to maximize the use of memory, CPUs, and disk space. 4GL programming languages did not exist so you actually had to write code to create those fancy pull down boxes in the UI. I am glad that we now have powerful computers and robust 4GL programming languages, but what we seemed to have left behind is systems design.
I can't count how many times I have reviewed a requirement document that did not specify how many transactions were expected, what the expected response time was, or what the required availability or uptime was. How can the developers even begin to design a system without knowing the technical requirements of the application?
I sometimes wonder if the majority of the developers raised in the "point and click" era really understand how the computer works under the covers. Don't get me wrong, I have seen many developers carelessly waste resources back in my mainframe days also, but it seems to be more common today.
Software engineering does not apply only to proprietary software. You still need to analyze and design even when implementing 3rd party software. How many failed projects have you seen with implementations of SAP or Peoplesoft? How many IT shops are afraid to take on initiatives like Enterprise Architecture, SOA, and Business Processing Reengineering? Many shops back away from anything large because so many of these types of projects fail. I believe they fail for two reasons, failure to manage change and inadequate design and planning.
So how did we get here? Is it because hardware is cheap and we can always "throw hardware" at our problems? Have we become so accustomed to dragging and dropping that we fail to think things out? Are we putting speed to market ahead of analysis and design or is software engineering becoming a lost art?
Amen brother...
As a profession we seem to have forgotten a lot of skills that IT practitioners learned back in the day. The things you mention are the kinds of skills that other professions like engineers, architects and doctors have built up over generations. They comprise the transferable body of knowledge that makes those professions as powerful and effective as they are today.
It's about time for the IT profession to get serious about creating a coherent body of knowledge that can be tranfered to and added to by the next generation.
I fully agree with you.
Part of the problem could be, that technology is evolving fast, but if you really look at it from a technical point, Ajax and all that new Web 2.0 "buzz" has been around for over 15 years, just that XML as "glue" was added to it lately. Same for SOA which now often fulfills, what Corba has failed on most attempts.
A bigger problem is, that people with little or no education (such as "Become a Rich IT Guru in just 4 Weeks" TV ads ;-) are entitled "Junior AVP Project Manager" and have simply no competence or idea what a project has to contain or a sound requirements documentation.
Not to mention, that they all abuse Agile for not being careful or coordinated enough in their efforts and Lean for Lazy ;-)
I'm not even sure to what degree the hardware implications of software engineering are discussed in education anymore. For web developers, most really do not have any formal training on the systems they are developing applications to run on. It's quite common these days for developers and IT to be silo'd. Only after an application gets rolled into production do the shortcomings of poor systems design come into play, and then we see developers and IT working together to fix the problem.
I think we have certainly grown accustomed to throwing hardware at the problem. It's hard not to as business requirements change throughout the development phase, forcing developers and engineers to change the scope of the project along the way. Deadlines are often favored over quality of work because you can always go back and fix the code later, when you have more time.
-Eric
I am with you. One of the problems is that technology and development move so fast that it is difficult to maintain a consistent transfer of knowledge from one generation of software developer/designer to another. With doctors, young internists learn from experience of seasoned MD's along with real world experience. With software, you don't have a bunch of web developers sitting down and absorbing the knowledge of a mainframe guy - in fact, most young developers view people on older systems as useless and have little to no respect for them. I once had a guy that graduated with a master's degree. First job, and I was the software designer. He looked at me as a dinosaur and argued EVERYTHING I said. Didn't matter that he was surfing to try and get ammo for his arguments. He never presented any correct or experienced info to support his argument.
Solid Software Enginerring practices should transfer to anything you are designing software wise and should be taught and practiced. But the dollar and the bottom line rule the day.
Very true, but on the projects I've dealt with, the kinds of requirements you mention belong in a non-functional requirements document (perf, scale, concurrent users, etc.) One problem I've seen frequently is that when technical info is included in the functional requirements, the writer tends to live in the database and does not have a holistic view of the system. So, you can end up with requirements that are very myopic and do not explain the business rationale at all. For example, I've seen reqs docs littered with items such as 'add X field to the database', but no explanation of why this is necessary, no reference to a conceptual business entity, no context whatsoever. Easy to fulfill technically, but at what longer-term cost?
I suppose all I'm saying is that I agree with the original post that both functional and technical requirements are necessary, but let's let the business people write the functional specs, the tech folks write the tech design and both collaborate on the non-functional to determine what's feasible given the constraints of the project.