Common Format and MIME Type for Comma-Separated Values (CSV) Files
draft-shafranovich-mime-csv-05

[Ballot discuss]
ABNF questions:

- VCHAR includes DQUOTE in the range, so

escaped = DQUOTE *( VCHAR / CR / LF / CRLF / 2*DQUOTE ) DQUOTE

could be '"""' (DQUOTE + %x22-as-VCHAR + DQUOTE)

- Spaces are not permitted in the grammar at all.  I don't think that's intended.

- Are non-quoted fields permitted to contain quotes, e.g.,
aaa,bbb,foo"bar,ccc
or does this have to be
aaa,bbb,"foo""bar",ccc ?

If spaces are allowed in unquoted strings, but quotes are not, then you could do

escaped = DQUOTE *( FOO / CR / LF / CRLF / 2DQUOTE ) DQUOTE
non-escaped = *FOO
FOO = %x20-21 / %x23-7E

where someone else can come up with a better name for FOO.  Since it's changing the definition from 2234's VCHAR it's probably not good to keep it named VCHAR.

...oops, and I just noticed that 2*DQUOTE also makes it ambiguous - that would allow "abc"""123", which doesn't fit the text description; I think it should just be 2DQUOTE.


[Re the other review mentioning COMMA inside escaped; VCHAR (and FOO) include COMMA]

[Ballot discuss]
ABNF questions:

- VCHAR includes DQUOTE in the range, so

escaped = DQUOTE *( VCHAR / CR / LF / CRLF / 2*DQUOTE ) DQUOTE

could be '"""' (DQUOTE + %x22-as-VCHAR + DQUOTE)

- Spaces are not permitted in the grammar at all.  I don't think that's intended.

- Are non-quoted fields permitted to contain quotes, e.g.,
aaa,bbb,foo"bar,ccc
or does this have to be
aaa,bbb,"foo""bar",ccc ?

If spaces are allowed in unquoted strings, but quotes are not, then you could do

escaped = DQUOTE *( FOO / CR / LF / CRLF / 2*DQUOTE ) DQUOTE
non-escaped = *FOO
FOO = %x20-21 / %x23-7E

where someone else can come up with a better name for FOO.  Since it's changing the definition from 2234's VCHAR it's probably not good to keep it named VCHAR.


[Re the other review mentioning COMMA inside escaped; VCHAR (and FOO) include COMMA]

[Ballot discuss]
How does the recipient know whether the header line will be
  present or not?  Section 2 says that the first line can be:

    field_name,field_name,field_name CRLF

  Perhaps a MIME parameter is needed to indicate the presence
  of this line.

[Ballot discuss]
Based on review by Joel Halpern:

This document is nearly ready for publication as an Informational RFC.
There is one small apparent BNF bug, one question, and one clarificaiton.

The BNF for 'escaped" does not list COMMA in the characters that can be included in the double-quoted string.  The early text (and common sense) indicates that comma needs to be able to appear there.  (Note also that the CRLF in the list of permitted characters is redundant given that CR and LF already appear.  Just change that to COMMA.)

I am not clear as to whether an Informational document can be used to create an IETF tree MIME type.  RFC 2048 says "RFC publication" which this clearly is.  It also says "requires standards-track processing" which this clearly is not.

Also, if the option of a header line is going to be described, then the text ought to indicate that recognition or expectation of such a header line is an application specific or local matter not covered by the format definition.  (There is no way for a computer to tell looking at CSV content according to these rules whether the first line is header or content.)

Yours,
Joel M. Halpern

[Ballot discuss]
Based on review by Joel Halpern:

This document is nearly ready for publication as an Informational RFC.
There is one small apparent BNF bug, one question, and one clarificaiton.

The BNF for 'escaped" does not list COMMA in the characters that can be included in the double-quoted string.  The early text (and common sense) indicates that comma needs to be able to appear there.  (Note also that the CRLF in the list of permitted characters is redundant given that CR and LF already appear.  Just change that to COMMA.)

I am not clear as to whether an Informational document can be used to create an IETF tree MIME type.  RFC 2048 says "RFC publication" which this clearly is.  It also says "requires standards-track processing" which this clearly is not.

Also, if the option of a header line is going to be described, then the text ought to indicate that recognition or expectation of such a header line is an application specific or local matter not covered by the format definition.  (There is no way for a computer to tell looking at CSV content according to these rules whether the first line is header or content.)

Yours,
Joel M. Halpern

AD evaluation comments: there are a few minor errors in the BNF.  The DQUOTE and VCHAR atoms are undefined.  VCAR is probably supposed to be VCHAR.  Please either define DQUOTE or import the definition from RFC 2234.