- From: Dave Kristol <dmk@research.bell-labs.com>
- Date: Mon, 3 Mar 97 09:58:54 EST
- To: koen@win.tue.nl
- Cc: http-wg@cuckoo.hpl.hp.com
> From: koen@win.tue.nl (Koen Holtman) > > [DMK] > >The messy part is for a SMG ("new cookie") aware server to distinguish an > >old from a new cookie, > > If I read the specs right: > > Old: Cookie: CUSTOMER=WILE_E_COYOTE > New: Cookie: $Version="1"; Customer="WILE_E_COYOTE"; $Path="/acme" > > Easy enough to distinguish: look for $Version at the start of the > header. Yes, but that's the easy case. It gets harder when you worry about multiple cookies in a Cookie header, choice of either ';' or ',' as a cookie (not attribute) separator, and the rules for stickiness for attributes. With a Cookie2 header, you can - dispense with $Version (for the default case) - require ',' as the cookie separator - do away with the attribute stickiness rules - dispense with reserving '$' as a special first character > > If I were to decode cookies, I would not even bother to distinguish > the formats beforehand. I would just code a parser which is able to > handle both formats. No debate there. > > I think that the spec would be messier, not cleaner, if a Cookie2 > header were introduced. I disagree, but you can judge for yourself. I've already told you where you can find a revised version of the spec that uses Cookie2: <http://portal.research.bell-labs.com/~dmk/cookie-2.36-2.42.txt> <http://portal.research.bell-labs.com/~dmk/cookie-2.36-2.42.ps> Dave
Received on Monday, 3 March 1997 07:05:36 UTC