[rfc-i] Relative attribute and the # character

Paul Hoffman paul.hoffman at vpnc.org
Wed May 27 08:05:22 PDT 2015


On the design team list, Julian brought up an issue with the new draft, and I thought it would be useful to move it here.

> On May 27, 2015, at 7:34 AM, Julian Reschke <julian.reschke at gmx.de> wrote:
> 
> On 2015-05-27 16:26, Paul Hoffman wrote:
>> On May 27, 2015, at 12:35 AM, Julian Reschke <julian.reschke at gmx.de> wrote:
>>>> 2.44.4.  'relative' attribute
>>>> 
>>>>   Specifies a relative reference from the URI in the target reference.
>>>>   This value is given without the leading "#" character found in
>>>>   relative references.
>>> 
>>> A relative reference gets resolved against a base URI. It does not necessarily start with a "#", so we can't just drop it because it makes the value ambiguous.
>>> 
>>> Example from RFC 7538:
>>> 
>>>> <xref target="HTML" section="4.2.5.3" relative="document-metadata.html#pragma-directives"/>
>>>> 
>>>> <reference anchor='HTML'
>>>>           target='http://www.w3.org/TR/2014/REC-html5-20141028/'>
>>>> ...
>>>> </reference>
>>> 
>>> Here, the section number 4.2.5.3 would be linked to <http://www.w3.org/TR/2014/REC-html5-20141028/document-metadata.html#pragma-directives>.
>> 
>> I thought the model for RFCs and Internet-Drafts that we were using would require a "#" and, if so, your desire would mean that making a relative link to an anchor in an RFC (not a section number) would require relative="#anchorname" instead of relative="anchorname". I'm fine with that if other folks are. It's pretty easy to describe that in the text.
> 
> Please.

Do folks here agree that, if they are going to specify an anchor for a relative link, they need to give it exactly (including the preceding "#" or "/")? That would make the feature more flexible, although it means people needing to understand how to do relative links better than they might. Personally, I think that restriction is OK.

--Paul Hoffman


More information about the rfc-interest mailing list