11.5: why submission/@separator?

Looking at section 11.5 of the current draft, I get the 
impression that the separator character for the 
application/x-www-form-urlencoded serialization is under
the control of the form author.

How much freedom does the author have in the choice of 
separator?  Is it restricted to either "&" or ";", or can 
it be chosen arbitrarily?  Section 3.3.3 indicates that
the default is ";", but doesn't indicate any other 
restrictions.

What concerns me is that a server-side parser might have
"trouble" figuring out the separator from the form submission
data.  For HTTP POST data, it's possible for the client to include
the separator within the Content-type header (just as the 
multipart/form-data encoding would).  However, this header isn't 
available for an HTTP GET.

IMO, leaving the choice of separator up to the forms author
is a mistake.  Most common "legacy" parsers already know
how to handle "&" and ";" as separators; I can't think of
any good reason why someone would need to use a different
character.

btw- "separator" is probably a misnomer here, since the 
draft's description (e.g. 11.5) indicates the character is 
being used as a "terminator".

-- 
Joe Schaefer

Received on Thursday, 5 September 2002 10:02:55 UTC