Child pornography blocking again

[Kraktus <kraktus@xxxxxxxxxxxxxx>]

Okay, here's my idea: Suppose exit servers included the term 'reject
cp_blacklist' in their torrcs.  If it worked well, it could be
included in the default exit policy.

cp_blacklist could be a variable rather than an explicit IP address or
hostname.  The IP addresses and hostnames included in that variable
could be known by the directory  authorities, so that it could be kept
up-to-date.

By making the blacklist directory information, Tor clients would only
have to download the list once each time they get the directory, not
separately for each exit server.

I realise, of course, there are problems with this.

* If the IP address is shared, there might be a child porn site on the
same IP as a legitimate site.
* hostname-based filters can be bypassed.
* User-contributed legitimate sites can sometimes have child porn on
them, until their management notices the material and takes it down.
While getting the management to take the material down should be
preferable to adding the site to the blacklist, the material will
still be available via the Tor network during the interim.
* Blacklists are never complete anyway.
* Definition of 'child porn'.  Perhaps separate variables, e.g.
cp_blacklist_strict, cp_blacklist_normal, cp_blacklist_conservative
could be used to deal with this somewhat?
* Backwards compatibility.  Maybe client recognition of variables
could be added a few versions before servers are given the ability to
use them?  One would also want the variables to be able to include
other variables, so, for example, cp_blacklist_normal would be
cp_blacklist_conservative plus a list of additional IPs and hostnames.
* Performance?

Is this idea even feasible?