Like anything, this depends on your data. If your data set contains a lot of commas, then using comma as a delimiter is not a very good idea. I generally use tab or whitespace delimited text files for quick and easy data input/output (which is part of the reason why tab-delimited text files are very popular in Bioinformatics).

Is there a specific class or type of logging that you're doing? The module Text::CSV_XS seems like a pretty good idea, especially if you're doing system level logging of network, user and filesystem accounting. The idea of using an XS module to do input and output of data formats is a good idea, and could easily be extended to use your delimiter of choice.