Q: (From Mr Peter Vanderbilt). Please comment on the use of route addresses as defined in RFC 822. Why did you not advocate them in your article? For example,
princeton! honey! down%purdue@csnet-relay.ARPA
could be phrased as
<@csnet-relay.arpa, @purdue.csnet, @princeton.uucp: down@ honey. uucp>
if I understand 822 correctly. It may not be a great work of art but at least it’s unambiguous.
Is the problem that many hosts cannot handle source routes? In this case, can purdue.csnet invoke the uucp mailer with the appropriate address (princeton!honey!down)? If so, please send me a copy of their sendmaiLcf, especially if they could generate a working return address (from honey .uucp)...
A: ... The answer to your ... question, as you have guessed, is that there are many hosts which do not understand source-routed 822 addresses. In fact, in the UUCP world, there are many hosts which do not understand any sort of RFC 822 message at all. Sending a message to UUCP-land means assuming that one or more links along the way will not understand any metacharacters other than ’!’.
As for the problem of constructing correct return addresses, this is a non-trivial problem, mostly solved on an ad hoc basis. I don’t know what Purdue uses in their ”sendmail.cf” file, and picked that particular example from Peter Honeyman’s paper (since he was kind enough to use an address that we had generated). In future I plan to construct a library of “correct” sendmail.cf files for distribution as samples to the membership; I’ll try to include one from Purdue at that time. In the meantime, I suggest you contact “postmaster@purdue.arpa” yourself if you are really eager for a sample of their file.
Comment: (From Mr Charles Hedrick). The reason why no one uses
@a,@b:c@d
is that it doesn’t accomplish anything. Host names a, b, and d must all be valid Internet host names. Thus this form cannot be used to express routes involving UUCP hosts. Since all Internet hosts can talk to each other directly anyway, this means that the syntax is of dubious usefulness. Most sites will optimize routes by delivering directly to the final destination. (That is, they don’t implement source routing.)
There was considerable discussion within the Internet community when this ruling was first made. Many of us felt that it was helpful to have a syntax to express routes to non-Intemet sites, and advocated that any syntactically valid host name should be allowed. I am not clear why the restriction was imposed.