Yes, that could be done. Problem is that the NSA's remailer(s) would immediately deliver messages to the destination. Get enough NSA remailers, and the web wouldn't be trustable. Now, remailers in the web can and should feel free to randomly forward mail to other remailers, but it's the sender who should pick the minimum chain length, and recursively encrypt their own envelopes.
Very good point. Still, I wish there was a way for my local software to automatically make this chain based on some sort of knowledge of what remailers are currently up. Ideally, my local software could figure out all this info without manual intervention on my part; it would maintain it's own list of remailers, and keep track of when they go down. I'm not sure it's possible to set up a system like this, but it would be enormously helpful. One naive solution would be for remailers to have a "ping" function. I could send a remailer a "ping" message, and it would just bounce some acknowledgement back. More likely, my software could do this periodically, and keep track of which remailers are down, or non existent, and not use those. The problem here is that an eavesdropper could get knowledge of which remailers I am planning on using, which could help traffic analysis enormously. The "ping" function could support anon encryption block, so that I can ping a remailer through several other remailers anonymously. This is an improvement, but the traffic generated by lots of people periodically doing this is going to be enormous. As it is in any implementation of this sort. [If you wanted to, you could make the remailers "ping" now by yourslef, just have a message resent to yourself. But we can't all do this automatically often, simply because of the traffic it woudl generate. I think.] The next idea I had involves a usenet newsgroup. Bear in mind I don't really know how this sort of thing works, so tell me when I've said something nit-witted. Anyhow, there could be an alt.remailer.net newsgroup. All participating remailers would post an "i'm here" message on it periodically, say once every 24 hours. This message would include the remailers public key as well. My local software could scan this newsgroup. If a remailer hadn't posted a "i'm here" message in 30 hours or so, my local software wouldn't include it in any chains. If it's been several weeks, my local software will drop it from my database of remailer's altogether. If a "i'm here" from a previously unknown remailer is found, my software adds it to the database. Or, if I'm worried about abuse, I only add it to the database if it's public key is singed by someone I trust. Okay, now everyone try to rip this plan apart. :) I'm sure I haven't arrived at the idea solution, but there's got to be some way to create a remailer-net that will allow my local software to generate long remailer chains to remailers that are all still existent (now, if one of the remailers included in my 6 remailer chain goes down, it's a major pain to figure out which one it was, and why my mail never arrived there), all automatically. Until we can arrive at such a system, remailers are never going to be really useful to a large number of people; it's just too generate secure remaielr ^?chains that are trustable.