In Windows, a file extensions are the part of a filename that comes after the last dot, for example, “.txt” or “.doc”. They serve to identify the kind of data in the file, so the OS knows what applications can open the file.
The Windows implementation of file extensions has been criticised on several counts: the mechanism has been extended and hacked several times to accommodate advances in technology, so it really is not very clean any more. However, these are implementation issues.
The real problem with the current scheme is the entanglement of the label that identifies the file and the type of information in the file.
Interestingly, in real life, the name of a thing often does indicate the type of of that thing. For example, “Marion” is likely to be female. “Ford” is likely to be a car. “Elvis” is likely to be a singer. On the other hand, this rule is not followed without exception. Is “Chris” male or female? “Marion Morrison” is male (aka. John Wayne), “Ford” is also a river crossing, “Elvis” could be the name of my cat (if I had one).
The solution is that, in principle, the name of a file and the file type should be separate attributes of the file. They should be independent part of the file’s meta-data, just like creation date, size, etc. Furthermore, the filename should be completely arbitrary.
For an example of the benefits, consider one of the most common problems with the current mechanism: some users think that if they change the file extension then they can open the file with a different application. Of course, this actually works for some files , e.g. “.log” -> “.txt” and “.config” -> “.xml”.In most cases, however, it fails. I have tried to explain to these users that expecting things to work this way is if calling me “Marion” would make me female, calling me “Ford” would make me a car manufacturer or calling me “Elvis” would give me the ability to sing (or miaow convincingly). Some users get this. Others do not. Of course, I could put this down to user stupidity. However, given that the whole issue could be avoided with better design, I prefer to think of this as the system’s stupidity.
The trouble is, it is a bit late to change things now…
All Wrong 
Marian Jones
I Googled for something completely different, but found your page…and have to say thanks. nice read.
I also found your site by googling something totally different.
I feel the way about folders. We should just use tags to identify files and locate them. Why force a file to be in a certain location and then have to remember which folder (of thousands) we stored it in. Tagging something as “Client X” as well as “Business Documents” would make it easier to locate files that pertain to multiple subject matters.
Brian B:
An interesting idea. I have been wondering about folders, too.
I do find folders useful sometimes, where information falls into a natural hierarchy. Those occasions may be rare, but where they exist, a nested structure can be useful as it provides additional information about the relationships between files.
Tagging has problems, however. What happens when you tag some files, “Business Documents”, and others “Work Docs”? I guess there must be solutions (auto-complete when entering tags, matching based on meaning, etc).
The other solution is just to rely on search – like Google does for the internet.
To be honest, I’m not entirely sure about the idea of files…
You’ve made me think, Brian. That’s always good.