ietf-nntp OVER extension

[Russ Allbery]

Clive D W Feather <clive at demon.net> writes:
> Andrew Gierth said:

>>  Clive>   X.X.2 The byte:count metadata item
>> 
>>  Clive>   The byte:count metadata item for an article is a decimal
>>  Clive>   integer.  It is the number of bytes in the entire article -
>>  Clive>   both headers, body, and separating blank line - except that
>>  Clive>   the US-ASCII CRLF at the end of each line counts as one
>>  Clive>   byte, not two.
>> 
>> Ew. No. Ideally, it should specify the byte count counting all CRLF
>> pairs as two bytes.

> That was my proposal six months ago, but everyone told me that existing
> practice is to treat CRLF as one byte.

Personally, I think existing practice is wrong, and since we're creating
something new, we can fix it.  The byte count should, so far as possible,
match the number of octets transferred on the wire when the article is
downloaded IMO (although it's arguable whether it should include the extra
dot from dot stuffing).

> Um, I can see the number of octets being different, because of dot
> stuffing, but I'm not sure I get the "isn't predictable" bit. Isn't the
> whole reason this is in the overview to *make* it predictable ?

Yeah, I don't understand that either.  Andrew, could you expand some
there?

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