Change Log

2006.02.22 Add the second script.
2002.02.04 First version

Raw Input		Desired Output
line 1 line 2 line 3 line 4 line 5 line 6 line 7 line 8 line 9 line 10		line 1 line 2 line 3 line 5 line 6 line 7 line 9 line 10

Script and Comments

Script₁

[ 1] :loop
[ 2] /^\([^\n]*\n\)\{3\}/!{
[ 3] N
[ 4] b loop
[ 5] }
[ 6] s/\n[^\n]*$//

Comments

Note:
If you use sed other than GNU sed version 3.02.80, you have to do a little modification, please take a look at Important Notes.
Line [1] thru [5] constitute a loop. This loop will repeat until the Pattern Space contains exactly 4 lines.
Step [6] will remove the 4th line. After this step, sed will output all the lines of PS and start a new cycle.

Script₂

[ 1] :loop
[ 2] N
[ 3] /\n[^\n]*\n/!b loop
[ 4] n
[ 5] d

Comments

This script use a different approach from the last script:

Action	Steps involved
Get the first line, keep it in PS.
Join next line, repeat this until PS has 3 lines.	Steps [1] thru [3] constitute such kind of loop.
Now PS has 3 lines, print them and read next line.	Step [4]
Delete it, since it is the fourth line.	Step [5]

This script assumes that when the last line of a file is reached, command 'N' will print the contents of PS then terminate.