On Fri, Sep 25, 2009 at 1:40 PM, Julian Reschke <julian.reschke@gmx.de>wrote: > Henrik Nordstrom wrote: > >> ons 2009-09-16 klockan 16:34 +1000 skrev Mark Nottingham: >> >>> So, strict BNF with a note in prose that old implementations may quote >>> other characters? Works for me. >>> >> >> Usually this kind of things is done the other way around. Open BNF with >> restrictions in prose on what may be produced. Or if one wants to go the >> long way by introducing "obsolete/old" BNF constructs for stuff which is >> should not be produced but still accepted by parsers for legacy reasons. >> >> Notes in prose should not require parsers to go outside the BNF. >> > > OK, for now I have only made the change disallowing non-HTAB control > characters in quoted-pair (< > http://trac.tools.ietf.org/wg/httpbis/trac/changeset/696>). > > I have not yet added any prose about recommending not to use quotes when > not needed. With respect to that, we need to decide: > > - whether that's purely advisory or a requirement (SHOULD?), I lean to the > former, and > > - where exactly to state it, as quoted-pair is used both inside > quoted-string and comment, and the characters that need escaping thus > differ; one way to fix this would be to change the ABNF so "comment" gets > it's own quoted-cpair rule. > I would say implementations SHOULD NOT use quotes where not required but MUST always accept them. Sam
Received on Friday, 25 September 2009 12:59:30 UTC