[ftpext] NAT64 client/server document

[Iljitsch van Beijnum <[e-mail]>]

Hi,

There was clear consensus (it wasn't even rough) yesterday that the current FTP64 document should be split up in an ALG spec and something else that specifies how clients and servers should behave to avoid the need for such an ALG in the presence of IPv6 to IPv4 translation.

The question is what would be the best way to approach this. On the one hand it's attractive to just update the standard. We can then obsolete PASV and PORT and mandate the use of EPSV and EPRT instead. I'm not sure that would accomplish much in practice, though, as many clients till issue PASV (or PORT) so servers will want to keep supporting that. And there's the issue of the boxes that work when there is a 227 response but not when there is a 229 response. I also worry that an overhaul of RFC 959 and rolling in stuff from later FTP RFCs would take quite a long time because all kinds of other stuff would find its way into the document, too.

The alternative would be to go for best current practice. This has the advantage that we can recommend some things that may or may not adhere to the strictest letter of the standard as long as doing them is obviously the best thing to do in practice. One of these things would be ignoring the address in 227 responses. On the other hand, BCPs carry less weight than standards, so people might find it even easier to ignore a BCP than a STD.

Thoughts?

The idea is that I will continue with draft-ietf-behave-ftp64 which will be reduced to the ALG spec and it looks like Dapeng Liu and myself will work together on the other document.

Iljitsch