Grep contains a lot of patches so that it is very fast even in UTF-8,
but only in the default case sensitive mode. If you type "grep -i
some_very_large_data" then you'll find that it is approx. a hundred times
slower than if you omit "-i". Example:

$ time grep foobar /usr/lib/locale/locale-archive

real    0m0.089s
user    0m0.044s
sys     0m0.048s
$ time grep -i foobar /usr/lib/locale/locale-archive

real    0m17.023s
user    0m16.977s
sys     0m0.044s

Note #1: I'm not using Fedora, but applied its patches (grep-2.5.1-52.2) to
grep 2.5.1a and recompiled it on my system and found this behavior there.

Note #2: If I apply the patches from SUSE's grep-2.5.1a-20 instead then it's
fast even in the case insensitive mode.

Note #3: "fgrep -i" is fast on Fedora, too.