On Sat, 22 Oct 1994 23:23:55 EST, Masataka Ohta said:
User agents are unrelated to the issue.
If a user uses encoding including NULL, his agent is designed to
accept NULL.
If not, regardless of how NULL is treated by a user agent, the
message containing NULL is wrongly encoded and unreadable.
Logic check here.
If I parsed this correctly, it's saying that if I have a user agent
that generates an outgoing NULL, it should accept inbound NULL as
well. This is all fine and good. Any user agent that can't receive
its own output is broken.
Unfortunately for Real World Interoperability, the actual problem is
that I can have a 100% <insert standard here> compliant MUA and MTA
that generate a NULL outbound. Now, in the Real World, people rarely
send e-mail to themselves. The problem arises when they send to some
OTHER user who has a 99.9% compliant MTA/MUA which botches on NULL.
Given the huge number of Unix systems running Sendmail, and the fact
that Sendmail wont support NULL correctly until the still-vaporware
8.7, and the fact that the standards AS WRITTEN explicitly require
NULL to be handled correctly, a large number of machines are only
99.9% compliant.
I'm having a non-drug-induced flashback about a voice chanting in the
night "tib 8 dneD tsuJ" - it's the same argument, but in reverse -
rather than breaking existing implementations by stepping outside the
standard, you're breaking existing implementation by complying.
Valdis Kletnieks
Computer Systems Engineer
Virginia Tech