fileinfo and mime types I've never heard of

Question

I'm not a stranger to mime types but this is strange. Normally, a text file would have been considered to be of text/plain mime but now, after implementing fileinfo, this type of file is now considered to be "text/x-pascal". I'm a little concerned because I need to be sure that I get the correct mime types set before allowing users to upload with it.

Is there a cheat sheet that will give me all of the "common" mimes as they are interpreted by fileinfo?

---

Sinan provided a link that lists all of the more common mimes. If you look at this list, you will see that a .txt file is of text/plain mime but in my case, a plain-jane text file is interpreted as text/pascal.

Answer 1

There is a chart that shows a list of common MIME types and their corresponding extensions. Here

Answer 2

fileinfo is a "best guess". It analyzes only a portion of the file in order to try to figure out what type the file is, and as such it can be fooled easily enough. Perhaps your file starts with a Pascal comment or keyword such as Project or Unit.

Answer 3

Fileinfo is not using the extension of the file to determine which mime-type it is, but (quoting) :

The functions in this module try to guess the content type and encoding of a file by looking for certain magic byte sequences at specific positions within the file.

The idea being that the name à of the file, and its extension, are provided by the users (especially in a case such as yours, where the files are being uploaded by users), and, as such, are less "sure" than the content of the file itself.

Maybe a solution could be to not check on the whole mime-type returned by fileinfo, but to only use the first part of it -- at least in some cases ?

For instance, maybe you could accept all mimetype that are in the text/* and image/* families, and refuse all those look like application/*, except for application/pdf ?
(Just an example -- but you see the point)

Answer 4

I have found that, as of at least version 5.03, the 'file' command can in some circumstances mis-identify a plain text file as a Pascal source file, simply because it contains the word 'program' or 'record'. At least that's how it looks having examined the source (src/names.h). I believe the php fileinfo command uses the same 'magic' engine, so I suspect this is the cause of the problem. If/when I am accepted on the file mailing list, I will notify the maintainers of this issue.