Ian Smith, Worcester, Yet Another Freelance Software Engineer/Programmer

I am an older programmer who started his working life writing assembler on a 1MHz 8 bit microprocessor with 256 bytes of RAM.

At the other end of the technology scale recently I wrote a C#, WinForms, ASP.Net, Web Forms, WCF system creating and fronting an SQL Server databases with 500 million plus rows, which was used to manage email marketing campaigns of millions of emails per month.

People who needed the power and flexibility had a desktop app, others who only wanted to see “results on an iPad” had a web app and some people who needed to access to the raw data has WCF web REST services (APIs).

In the middle size wise, I wrote a full cycle EDI administration and quotation system for household and motor insurance using MS Access/VBA to provide a front end to an SQL Server database and a C++ quotation engine.

The important point of mentioning these systems is that they show that a lot can be achieved using a variety of tools and that I am happy to go with the ones that you want.

There is of course the caveat that the choice must make a reasonable amount of sense.

The assembler background is important as it teaches a programmer an awful lot about what has to go on to achieve anything even vaguely useful.

So when you write a high level command you have an understanding of the computing resources that you are asking for and it makes you think. Should I really use that library function or write some code that will achieve the same thing using far fewer resources but take a lot longer write?

The modern trend is the opposite off this, it uses design patterns things like MVC (model/view/controllers) and javascript libraries (Angular, React etc) to form the basis of a system because you don't trust your staff to do the job properly.

The problem with these patterns and libraries is that they have apparently sensible roots and they let some people in IT departments feel superior to others as they understand them and why they are good.

it's only once you get deep into your project they you realise that they leave you restricted in how you develop your applications and often result in fragmented and difficult to maintain source code and are often resource hungry.

All to often they also leave the project owner stuck 10 years down the line when they are abandoned by their creators in favour of a new "best way" of doing things.

And whilst neither PerBI or IPL have been commercial successes they are complete standalone Visual C++ applications that exist in the market place. One where users can send in emails saying that It doesn't work, please fix it.

C++ is an evolution of the C language, for some tasks it is an improvement and for other tasks the new features are off no benefit.

Most people who are familiar with C/C++ are also familiar with VB/VBA/C# but many people familiar with VB/VBA/C# are not familiar with C/C++.

The principle advantage of C/C++ is that the language doesn't attempt to interfere with what the programmer has asked to do, or to provide slow generic solutions to problems.

This tends to mean that C/C++ programs run faster and are much smaller than programs written in other languages, depending upon the task these are still important factors

The really big downside of C/C++ is that they are the hardest languages to learn and the easiest language in which to write buggy programs, the programmer has to think of everything that might go wrong and remember to return to the Operating System any resources, such as memory, that it has asked for.

In this case the objective was to deskill the maintenance of motor insurance quotation engines which were initially written in C++ using COM to interact with an administration system created in MS Access/VBA.

The first step is a break the processing up into steps, here calculate the base premium.
Each step is composed of commands each of which have a dialog box ensuring that a well formed command is created.
Once all of the steps are completed each of these steps are then converted to C++ code and merged with boiler plate code to makes up a full quoter.

Finally An NMAKE project is automatically created and invoked to end up with a dll that can be passed to the live systems.
This is the final C++ program.

You can see that some of the C++ code directly maps to the commands in the Access form but aso that some doesn't. The code that doesn't is necessary for the quoter to work but never changes and the person maintaining the quoter doesn't need to know about it.

For example the assignment to q_FactorSave makes the result of this step available via a COM method .

Indeed some programming languages use this increase in power to increase programmer productivity at the expense of slower running programs.

After all if the program produces a result in 1/1,000 of a second rather than 1/1,000,000 of second who cares?

The problem is that as the amount of data that needs to be processed increases this becomes 1 second or 1,000 seconds and then everyone cares.

The reality is usually not this extreme, the time increases to almost a second, then a second or two and then even more. During this period the users gets fed up with the system but the delay is not so great that the time to fix the issue can be justified.

And this is not as farfetched as it might sound, but all it means is that that you should change the programming language that you use for this resource intensive step. Or possibly discard a built in library function of that language and replace it with one written for your specific case

Quite often this slow down was predictable if the programmers had understood exactly what they were doing, but It worked fine with our test data of 10 customers. .

This "simple" replacement is often impossible in reality as your programmers may not have necessary skills.

The explosion of the internet hasn't helped in this area, people learn some PHP/ASP.Net and decide that they are programmers and don't realise how much they don't know.

Equally their managers also don't realise how much they don't know and make decisions based on their advice.

There are already many people specializing in brochure or e-commerce sites.

Nowadays most websites are a commodity product, your site creator will pull elements from their "stock" and customise them to your requirements.

This means that the best prices with the most choice will usually come from people who specialise in web sites and have a very large library to draw on.

As I don't have that library I would have to charge you the full cost of originating the work, rather than spreading it over say 20 clients.

So if what you want is pretty much a copy of something that has already be done then the best place to get it from is from the people who created the original.

If however what you want is technically complex and visually simple then we may be a good match.

For example a site that needs a unique and complex database design with say 100,000,000 rows or one that does complex calculations but gives relatively simple answers.

Many web sites involve little or no programming at all, so a company that can set you up with a site that has a comments section and a CMS that allows you to post articles may be at their limits programming wise.

To try and explain this difference I put up the car insurance price comparison site example. The more complicated programming bit is the process that calculates the premiums, something that you don't really see, only the results.

We often see on the news that a big government IT project has failed and may think that I could have done that along with a few people that I know and in a few months.

The assumption is nearly always that it is the programmers or the system's designer that got it wrong, this is convenient but is it fair?

Or was it the case that the project would have been a failure even if were 100% complete and fully met the specification?

My worst project was one that should have been 6 month's work but it got bogged down because the commissioner of the project wanted a new system and the rest of the company wanted to keep the old one.

Another project used SSIS/SSRS when a C#/VB application was more appropriate because the project commissioner understood SSIS/SSRS and not C#/VB and had only gone to an outside supplier because he was being pressured internally to get the project completed.

So it is very important to agree not only what the objectives are but even if they are achieved will the project be a success?