Media Type "text/csv": new draft (-02) and Last Call

[Graham Klyne]

At 14:08 30/03/05 -0500, Yakov Shafranovich wrote:
>clyde.ingram at edl.uk.eds.com wrote:
>>But the same cannot be said of automated computer-based applications, 
>>where maintaining a strict count of generated and expected 
>>Comma-Separated-Values per record is not only easy, but also allows for 
>>an extra level of data validation: namely that a received record is 
>>corrupt if it has too few or too many fields.  This is where 
>>standardisation in the format of the CSV records becomes appropriate 
>>material for an RFC.
>
>The draft as it is written now (-03) does not mandate that the same number 
>of fields need to appear on each line, mainly due to the fact that the 
>draft is focusing on the MIME type registration. Would the following 
>change to section 2, subsection 4 be sufficient to address your concerns:
>
>"Within the header and each record there may be one or more fields, 
>separated by commas. Each line should contain the same number of fields 
>throughout the file. The last field in the record may not be followed by a 
>comma. For example:"

I think this change would *decrease* the utility of the proposed MIME type 
registration, which I see as being to document existing practice so that 
new applications can interoperate with it.  I think the benefits of 
interworking with existing practice greatly outweigh the purported 
advantage of (weak) integrity checking.  In this case, I really think that 
less [specification] is more.

#g


------------
Graham Klyne
For email:
http://www.ninebynine.org/#Contact