- device is mobile or not
The purpose of the data is to create various statistics later. So it's not important to identify 100% of the user agents correctly. What is important:
- identify ~99% of the cases correctly,
- clearly identify the cases where we couldn't identify the properties (this is important, if the unidentifiable percentage grows above 1% then we should improve the identification),
- use as little time and memory as possible,
- be simple to implement :-)
I have never worked with user-agent strings before and now I a realized that user-agent string are quite complicated.
I found the browscap project and I found a java library on top of that: https://github.com/blueconic/browscap-java
This seems good, but I am worried about the complexity and sheer size of the CSV browscap-java is using: it's 200Mb.
So I'd like to ask if this (browscap-java on top browscap) is the sensible tool for it or not?
Are there any alternatives, recommendations?