There is more than one way to do it

Quite often we need to do something (system engineering, software development, database,…) and when we let ten people do it, it will be done in eleven different ways.

Some organization like to „standardize“ things…

And to some extent that is needed actually.

But the challenge is to find what needs to be standardized and what not.

While it is absolutely clear that it will lead to chaos, if nothing is standardized, because stuff will not work together well and people will have a hard time taking over somebody else’s work or helping each other.

But the other extreme can be bad also… It kills creativity, it forces people to use inferior solutions that need a long time to implement. Just think of standardizing on hammers and then using screws as nails. Also to much standardization freezes an approach too early, when there should be different approaches competing, with the more successful ones gaining terrain.

What can be done and what works: for a big software it is possible, that different teams work on different components. They can use different programming languages. In some cases they should actually do that, if they need specific strengths of one language in one area and of another one in another area. Of course a lot of things like interfaces, security, character encodings, networking and network architecture need to be standardized. While it may or may not be perfectly legitimate ot use for example both PostgreSQL and MongoDB and CassandraDB by the same application, there needs to be a good reason for using two different relational databases.

I have seen this happen, but at some point of time one team using programming language A was quite successful and team using programming language B not. And there was no strong specific benefit of language B for what they were doing, other that they knew programming language B better. Companies can replace teams or train people on language A. Or try to train them and replace some of them depending on the success of the transition. Possibly it is not the language, but the team that was the reason… Not necessarily the people, but for some reason they got on the wrong track and are unsuccessful this time… In the end, the project decided to standardize on language A and have this part rewritten by people who knew language A quite well. Even though this meant writing stuff twice, in the end it brought a much better result. This „rewriting something with better technology“ can be a dangerous trap. Often the second solution takes ten times longer to write until it does what the „inferior old solution“ did. But in this case that problem did not occur at all.

Now we should look at the IT work from a more abstract point of view.

Some work is done and in the end there is a result. This can be a source code, a running software, a server setup, a document, a database setup, a network setup,…

As a general rule, we want to standardize some aspects of the result. How the result is achieved should be left to the engineer.

So for many ad-hoc-tasks people can use different tools, like different editors and IDEs, different configurations for these, scripts in different scripting languages etc. Of course efficiency is a goal. Scripts can even be documented or checked into git for optional usage by those who like to use them. People can use Perl, Python, Ruby, Bash or something else for scripting. But the scripts can actually become results and then become candidates for standardization. For example, for managing a large number of servers, a lot can be done using ansible, which can be seen as a kind of fourth generation (declarative) language for doing exactly this. It makes sense to agree on using this all over the place and then the ansible playbook and not only the result of its execution is the result of the work and probably checked into git.

When working on Linux-servers, vi or most likely vim will be installed on every server. With the default configuration, at least for root and other system users. It is possible for an organization to agree to install a second editor on every server without too much effort. But it is usually not done and it does not really solve the issue, so most of the time editing on servers is standardized to usage of vi. Get used to it, it will be like that anyway in the next organization you are working for.

When someone writes a super complex script that nobody else understands, then this can be a combination of unnecessary complexity and inherent complexity. If the inherent complexity is high, then it is quite likely that this script will and should be reused and then it is no longer an ad-hoc solution but it becomes an result. Then it becomes of interest to make it a bit maintainable and readable, so others can work with it and improve it.

But in all levels there are often different ways to do something that are all legitimate and good and efficient for the purpose.

Share Button

Schreibe einen Kommentar

Deine E-Mail-Adresse wird nicht veröffentlicht. Erforderliche Felder sind mit * markiert.