Even today where the Internet speaks UTF-8, there are still a lot of data out there that consists of printable ASCII characters only. One famous example are URLs (URIs, URNs).

These data are often human-readable, which is often a good idea but sometimes you might want to hide e.g. URL paths and query parameters a bit for too curious eyes. Encodings like Base64 add 33% overhead which can be done better when the input data are not arbitrary binary data but printable ASCII caracters only.

There are 95 printable ASCII characters. If we map all other byte values to one distinct "error" value, we have 96 different characters. A string of several consecutive characters can be encoded in a base-96 number, so 9 printable ASCII characters can build a number up to (96^9)-1 = 69'253'399'582'480'255. This number can be expressed in ASCII letters and digits, hence a max. 10 digit base-62 number. So this encoding has only 11% overhad.

To make things a bit more complicated I also added a little quirk to the codec: A 64 bit "secret" which scrambles the encoding a bit and without knowing this secret it is not so easy (albeit not impossible) to decode the data.

The name "base-96-to-base-62" would be too long, so I chose the shorter name "basilikum", which just means basil (Ocimum basilicum) in German.