Long file names can be a serious pain... for copy and backup reasons.
I recall a NTG client that used a long name for the directories,.. and it made updating nearly impossible Should follow the KISS method... Directory names don't need to be 120 characters long,.. and four deep.
So, do you think the reason I am seeing a lot more gzip in use with tarballs is due to the familiarity of gzip and the negligible difference in the compression between it and bzip2? Basically, bzip2 doesn't make enough of an improvement with sufficient regularity to entice people to move away from gzip, or is there some other benefit to gzip that my training material hasn't covered?
That's correct. The difference between the two is generally small enough that people are not concerned. And lots of systems still don't have bzip2 installed by default so if you want scripts or whatever to work universally you often use gzip because you know that it is always there and predictable.
Sorry if this was mentioned but I didn't see it directly mentioned for clarity:
If your compression is unavailable directly in tar (-J being essentially 7zip, my favourite), you can tar it first (without compression) and then compress the tar, this maintains both Unix metadata and also gives the benefit.
Also, if you compress something already compressed you won't get the best benefit, at least not when it comes to using something as powerful as LZMA/7zip.
I believe that that is mentioned in the tar article.