This is mainly a description of gnu sed (2.05) bugs, which should be the most used sed around. plus sides of gnu sed (2.05) . no limits on label size . no limits on buffer size . no limits on RE size (although REs may actually crack down) . freely available . support for 'cmd1; cmd2' . support for comments '#...' down sides: . very slow in respect with other seds . some irritating bugs . re lib too big and slow . no documentation ===== bugs desc: 1. substitution flag `t' command should perform a jump, only if some s/// has suceeded on previous commands (and the flag has not been reseted yet) gnu sed crack down when a number suffix is added, things like s/./&/20 t foo should jump to foo, iff there is a 20th char on the line (read: pattern space) gnu sed actually will always do the jump, on non empty lines in other words: gnu sed will see if it can make the first change, and on true, triggers the flag; this is wrong if a count is supplied 2. beginning white space sed should ignore all leading white space, in any context. to actually include leading white space, a slash is introduced : /foo/ { i\ bar } should outputs =>bar =>...some line with foo... gnu sed outputs => bar =>...some line with foo... 3. \n inside [] there is no easy way to include \n inside [char set] because `\n' is seen as a `n' which is hardly understandble the only current way, is to make a: s/[^\ ]/replacement/ 4. \n inside y/// making this simple: there is no way to make y command work no a newline people tend to use the s/// command for that, which is inneficient 5. lin1,lin2 b