[Top] [All Lists]

Re: Question related to draft-hansen-4468upd-mailesc-registry-02.txt and RFC3463

2007-08-09 07:44:27

Hi Ned,

Translating ESCs to readable plaintext is an exciting proposition but only 
if a message or other user interface containing the translation does not 
lose the original plaintext portion in the original DSN report without the 
user's consent.  That's simply a courtesy and, while I'd like to think 
ESCs can convey every possibility with good enough gradation to be useful, 
obviously there's the possibility that the code won't be a substitute for 
the human-readable supplemental lines provided by the sending MTA.  
Whether that's important to a non-English user is obviously hard to say, 
but it wouldn't be a problem in any event if every MUA actually did 
translation of the machine-readable portion for itself.  Since (as far as 
I can see) they do not and the technique is mostly reserved by automatic 
robots (MLMs, etc), the only question now remaining is how we might let 
MTAs rewrite DSNs on behalf of our non-English-speaking friends who use 
MUAs that will not translate.

For me, I'm thinking the answer is that the MTA delivering my mail send 
not only the DSN to my mailbox but also a rewritten version of DSNs it 
transports for delivery to local users as detected by a null sender and 
multipart/report parts; the text part of the report can tell me about the 
original copy I've been sent (multipart/report MIME parameters telling of 
the original Message-Id, perhaps, and how this is only a translation?) and 
contains an enumeration and translation of status in the original's 
machine-readable section.  The machine-readable section could, 
furthermore, also be rewritten such that parenthetic text be translated to 
a local language.  The resulting message should be a valid DSN in and of 
itself and, if local users chose, could be the only version of a DSN 
received (the original being discarded - although perhaps there are 
dangers there).  Also, this service is more-or-less exclusively useful to 
users reading mail with user agents, since the criticality of multiple 
DSNs for a single transaction being misinterpreted by a computer program 
to mean multiple identical failures (EG: MLMs seeing two failed sends when 
only one is intended) might be significant.  There would have to be some 
breathing room somewhere in the spec to avoid that possibility if this 
sort of thing were to be preferable.  Alternatively, if it isn't actually 
necessary to preserve the machine-readability of translations, then 
breaking the DSN format isn't an issue and the problem goes away.  The 
point is to have machine-assisted translation but only if the kind of 
rewriting done isn't destructive, and I think a separate message is the 
best way, although MIME encapsulation is just as likely to work.

As a sidenote: sendmail seems to do a limited amount of translation on 
SMTP responses containing ESCs.  You can find them in DSN transcripts in 
the plaintext portion of the report.  The translations are not thorough 
and are based as much on the code as on the state sendmail's client code 
was in when it hit trouble.


Sabahattin Gucukoglu <mail<at>sabahattin<dash>gucukoglu<dot>com>
Address harvesters, snag this: feedme(_at_)yamta(_dot_)org
Phone: +44 20 88008915
Mobile: +44 7986 053399