[xml2rfc] References feature

julian.reschke at gmx.de (Julian Reschke) Sat, 01 July 2006 00:16 UTC

From: julian.reschke at gmx.de (Julian Reschke)
Date: Sat, 01 Jul 2006 00:16:05 +0000
Subject: [xml2rfc] References feature
In-Reply-To: <8537.1151705802.245982@peirce.dave.cridland.net>
References: <01M48MPYZAVW0008CX@mauve.mrochek.com> <Pine.LNX.4.10.10606301443160.6820-100000@shell4.bayarea.net> <01M48NPH8U2I0008CX@mauve.mrochek.com> <8537.1151705802.245982@peirce.dave.cridland.net>
Message-ID: <44A62123.2070006@gmx.de>
X-Date: Sat Jul 1 00:16:05 2006

Dave Cridland schrieb:
> No, you don't have to - the idea would be that if you removed the only 
> normative xref (ie, <xref ... thing='normative'/>) leaving only 
> informative ones (ie, <xref ... thing='informative'/>) then it'd 
> automagically happen. No wading required - I'm sure an amusing reference 
> to Dallas can be made here.
> 
> The actual reference elements would be thrown in a single container, and 
> not marked normative or informative, xml2rfc figures out the 
> normativeness of them by the presence of a normative xref. Any others 
> get turned into informative references.
> 
> Whether or not this is worth doing (or, just as important, worth 
> changing xml2rfc etc for) is a different matter, and personally I'm 
> undecided, but I'd like to be sure that you're against it for the right 
> reasons. :-)

My 2 cents (from an xml2rfc implementation point of view):

A change like that would make the processor more complex (more special 
cases and edge cases to be tested). The current (historically grown) 
model where the processor needs to special case the "old" format (only 
one references section) and the "new" format (potentially multiple 
references sections with optional titles that need to be combined into a 
single document section) is bad enough. Furthermore, that kind of 
automatism seems to be handy only when types of references are 
frequently changed.

That being said, I *do* see value in optionally marking up the type of a 
reference inside the reference itself. However, that shouldn't cause any 
document output change. But it *could* be used by the processor for 
issuing warnings (if a reference is cited as normative, but doesn't 
appear under Normative References).

Also, a processor that knows about BCP14 keywords (see 
<http://greenbytes.de/tech/webdav/rfc2629xslt/rfc2629xslt.html#ext.element.bcp14>) 
could start to warn about text that combines normative requirements with 
  informative references.

Best regards, Julian