Re: FWS problem

["Charles Lindsey" <chl@xxxxxxxxxxxxxxxx>]

In <41BD7D35.7239@xxxxxxxxxxxxxxxxx> Frank Ellermann <nobody@xxxxxxxxxxxxxxxxx> writes:

>Charles Lindsey wrote:
> 
>> This specification uses the terms "header", "header name",
>> and "header content" which are synonymous with the [RFC2822]
>> terms "header field", "field name", and "field body" 
>> respectively.

>For your MUST you either need "header content = field name
>plus body", or simply use "header" (if that's name plus body):

No, the whole point of "header content" is that it does not include the
header name (if you want to refer to the header inclusing its name, then
we already have the term "header" for that). The question at issue is how
much of the FWS (including the obligatory SP) is to be included in the
term.

>> The header content of every header line (including the first
>> and any that are subsequently folded) MUST contain at least
>> one non-whitespace character.

>Yes, that's the erroneous paragraph.  You could replace it by

No, that is definitely the effect we want, given the intended meaning of
"header content".


>>> why this fuzz about a mandatory SP ?
> 
>> Because much deployed software breaks without it. ANd for
>> the same reason it is also required by the new NNTP draft.

>Please confirm, you're saying that "name: SP CRLF SP body CRLF"
>is okay, but "name: CRLF SP body CRLF" breaks some obscure NNTP
>servers ?  I have great difficulties to believe this, because
>for me it's "obvious" that trailing white space is always evil.

Things that will certainly break existing software are:

name: body CRLF
name: CRLF

It may well be that the case you mention "name: SP CRLF SP body CRLF" will
not actually break anything, but OTOH some "helpful" intermediate site
that thinks trailing white space is evil might well change it to
"name: CRLF SP body CRL", at which point it _will_ break existing
software.

So that is why we forbid empty content on the first line. RFC 2822 also
forbids enpty content on folded lines for essentially the same reason,
namely that the line "SP CRLF" might well be changed to "CRLF" by that
"helpful" intermediate site, whereupon it becomes indistinguishable from
the divede between headers and body.


>Is this NNTP draft something we can kill ?  The 2822 style of
>folding is much better than a mandatory trailing white space.

No. But the only point about it is that it requires the SP after the
name:. It does not concern itself with folding.

>Only the 2822 obs-style "name:body CRLF" is bad, but a simple
>"name: CRLF SP body CRLF" should be okay.  And all variants
>with TAB instead of SP of course.

I think some existing software also breaks with

name: TAB body CRLF

which is why we insist on SP.

-- 
Charles H. Lindsey ---------At Home, doing my own thing------------------------
Tel: +44 161 436 6131 Fax: +44 161 436 6133   Web: http://www.cs.man.ac.uk/~chl
Email: chl@xxxxxxxxxxxxxxxx      Snail: 5 Clerewood Ave, CHEADLE, SK8 3JU, U.K.
PGP: 2C15F1A9      Fingerprint: 73 6D C2 51 93 A0 01 E7 65 E8 64 7E 14 A4 AB A5