The [ICMP Echo protocol][1] (usually known as "Ping") is mostly harmless. Its main security-related issues are:

 - In the presence of requests with a fake source address ("spoofing"), they can make a target machine send relatively large packets to another host. Note that a Ping response is not substantially larger than the corresponding request, so there is no multiplier effect there: it will not give extra power to the attacker in the context of a denial of service attack. It might protect the attacker against identification, though.

 - Honored Ping request can yield information about the internal structure of a network. This is not relevant to publicly visible servers, though, since those are already publicly visible.

 - There use to be security holes in some widespread TCP/IP implementations, where a malformed Ping request could crash a machine (the ["ping of death"][2]). But these were duly patched during the previous century, and are no longer a concern.

It is common practice to disable or block Ping on publicly visible servers -- but being _common_ is not the same as being _recommended_. `www.google.com` responds to Ping requests; `www.microsoft.com` does not. Personally, I would recommend letting all ICMP pass for publicly visible servers.

Some ICMP packet types **MUST NOT** be blocked, in particular the "destination unreachable" ICMP message, because blocking that one breaks [path MTU discovery][3], symptoms being that DSL users (behind a PPPoE layer which restricts MTU to 1492 bytes) cannot access Web sites which block those packets (unless they use the Web proxy provided by their ISP).


  [1]: http://tools.ietf.org/html/rfc792
  [2]: http://en.wikipedia.org/wiki/Ping_of_death
  [3]: http://en.wikipedia.org/wiki/Path_MTU_Discovery