"If you are packaging ed for a distribution, please, try to use the lzipped source tarball, as this can improve the support for the lzip format in packaging systems. Thanks."
Great. Now we have .xz, .lzip (.lz) and .lzma. Not to mention .7z - all using basically the same compression engine, with just enough differences to make them incompatible with each other.
-a does work with decompression, what version are you using? I'm using GNU tar 1.26, and just successfully used "tar xaf firefox-9.0.1.source.tar.bz2".
If you explicitly provide a file by path, then tar doesn't use magic number autodetect, but uses file extension dispatch instead. You would need to pass the data as a stream to force magic number autodetect:
<firefox-9.0.1.source.tar.bz2 tar xaf -
Does that still work? It doesn't for me.
I was using (GNU tar) 1.20 and Debian stable is now at 1.23, but I think that autodetect code went in around 1.18.
The maintainer of GNU ed is also the author of lzip.
Though lzip predates xz, this GNU ed release is the first .lz file I've ever seen in the wild. Everyone else that distributes .tar.gz files (compatibility) along with .tar.something_else files (higher compression) seems to use bzip2 or xz, not lzip.
I've had to resort to ed a few times to rescue a system with no /usr partition. 'vi' or 'vim' happened to be in /usr/bin, so I needed to do some quick editing. Ed is definitely worth learning!
It's always good for systems administrators to be able to do things with one (or both) hands tied behind their backs. Murphy always seems to conspire to create a situation where those skills are necessary....
I actually found studying ed (while reading _The Unix Programming Environment_ by Kernighan and Pike) to be really helpful in understanding sed and vim in more depth.
If you assume you'll never need the 8th bit of any given character (which you wouldn't, if you only ever handle pure 7-bit ASCII) you can safely use it for storing a single bit of per-character metadata to avoid consuming excess memory.
GNU has a long history of ignoring POSIX where they disagree with it. There's a much better chance that the BSD implementation will be fully compliant.
Great. Now we have .xz, .lzip (.lz) and .lzma. Not to mention .7z - all using basically the same compression engine, with just enough differences to make them incompatible with each other.