ietf-nntp NNTP and 16-bit charsets

[Russ Allbery]

Clive D W Feather <clive at demon.net> writes:
> Russ Allbery said:

>> nnrpd rejects articles with lines >1024 characters; this is also a bug
>> in INN that needs to get fixed (I didn't realize that it had this
>> problem until fairly recently).  Apart from that, in transit, INN
>> handles unlimited length lines in the article body, and so do all the
>> other servers I'm aware of.  Is this provision really needed?

> I was trying to be conservative.

I'm wholeheartedly in favor, and much appreciate the summary.  :)  Sorry
if my tone didn't communicate that particularly well.

> What does the majority think ?

Yes, speak up folks.  :)

One point to be aware of in setting this limit is that historically news
has followed mail, and RFC 2822 states:

   - Lines of characters in the body MUST be limited to 998 characters,
     and SHOULD be limited to 78 characters, excluding the CRLF.

What this means in practice is that news articles with lines over 998
characters may not be gatewayable into mail without adding a transfer
encoding.  I don't think this argues against accepting them in news, but a
warning there may be worthwhile.

I think the limit should match mail; it should be 998 octets if there's a
limit.

Note that USEFOR currently requires relaying agents to support lines of
unlimited length (at the MUST level), which would imply that relaying
agents are required to apply some sort of transparent and reversible
content encoding to deal with longer lines if they have a lower limit.  I
think that's a bit of the tail wagging the dog; the limit should be
relaxed in the primary transport protocol first and then it can be relaxed
in the message format description.

RFC 977 punts to RFC 850, which punts to RFC 822.  RFC 1036 (which
replaced RFC 850) similarly punts to RFC 822.  So I think RFC 2822 is the
current canonical reference for RFC 977 purposes, and it says 998 octets.

So after further research, this does appear to be a change, and therefore
may be on shakier ground than I'd thought.

-- 
Russ Allbery (rra at stanford.edu)             <http://www.eyrie.org/~eagle/>