W3C home > Mailing lists > Public > public-html@w3.org > August 2009

Re: The cite and pubdate attributes

From: Ian Hickson <ian@hixie.ch>
Date: Tue, 25 Aug 2009 00:56:18 +0000 (UTC)
To: Jeremy Keith <jeremy@adactio.com>, Leif Halvard Silli <lhs@malform.no>, Henri Sivonen <hsivonen@iki.fi>, Jim Jewett <jimjjewett@gmail.com>, Bruce Lawson <brucel@opera.com>
Cc: HTML WG Public List <public-html@w3.org>
Message-ID: <Pine.LNX.4.62.0908250037310.13789@hixie.dreamhostps.com>
On Sat, 15 Aug 2009, Jeremy Keith wrote:
> Hixie wrote:
> > pubdate="" on <article> is solving the problem that there's no other way
> > to associate a publication date with a blog entry in HTML, in particular
> > to allow for conversion to Atom.
> 
> There is one other way. The hAtom microformat uses class="published" to 
> indicate the date(time) that an article was published (specifically, the 
> containing element with class="hentry").

The whole point here is to be able to convert HTML to Atom without having 
to annotate the document with classes everywhere, so a solution that 
relies on sprinkling classes everywhere is a non-starter. :-)


> Would it be too fragile to attempt the same kind of scoping that hAtom 
> achieves (by using "published" nested in "hentry") using <header> and/or 
> <footer>?

On Tue, 18 Aug 2009, Jim Jewett wrote:
> 
> Too fragile.
> 
> <header>
> <h1>My Brother's Wedding:
>     <time datetime="2008-07-04">4th of July 2008</time></h1>
> <p>Pictures of the bridal party and the reception fireworks</p>
> </header>

On Wed, 19 Aug 2009, Jeremy Keith wrote:
> 
> I guess I've still got my microformats thinking cap on where solving 80% 
> of the use cases is considered a success. In a markup spec, that really 
> needs to be close to 100%, doesn't it?

No, not at all. We're only trying to do the 80% here too.


> What if @pubdate were a Boolean attribute to be applied to a TIME 
> element rather than a name/value pair applied to ARTICLE?
> 
> So instead of....
> 
> <article pubdate="2009-07-30">
>   <header>
>     <h2><a href="blah" rel="bookmark">Accessibility of HTML5 video</a></h2>
>     <time datetime="2009-07-30">Thursday 30 July 2009</time>
>   </header>
>   <p>Brilliantly witty, incisive prose, in a gloriously elegiac style
>   reminiscent of <cite>Cider With Rosie</cite>.</p>
> </article>
> 
> ....we get:
> 
> <article>
>   <header>
>     <h2><a href="blah" rel="bookmark">Accessibility of HTML5 video</a></h2>
>     <time datetime="2009-07-30" pubdate>Thursday 30 July 2009</time>
>   </header>
>   <p>Brilliantly witty, incisive prose, in a gloriously elegiac style
>   reminiscent of <cite>Cider With Rosie</cite>.</p>
> </article>

I don't really see how that's any better. The data is still duplicated, 
the data is still hidden, and now we've just added even more ways to get 
the errors in the data (e.g. what if there are two conflicting <time 
pubdate> elements?).


On Sat, 15 Aug 2009, Leif Halvard Silli wrote:
> > 
> > I've removed cite="" from <section> and <article>, since their use 
> > cases are either weak or solved by other problems.
> 
> What does "solved by other problems" mean?

I meant solved by other solutions, sorry. :-)


> If it was meant "solved by other means", which means? Microdata?

rel=bookmark, primarily.


> > I've not removed pubdate="" since I don't see how else to get an 
> > unambiguous date out for conversion to Atom without annointing a class 
> > name.
> 
> I don't know why a @class ought to be better. But a predefined class 
> name solution should in principle be possible to introduce via the 
> profile/versioning *concept*. (HTML 4 embeds a default profile, HTML 5 
> embeds another profile.)

We originally had something like this, but predefined class names turned 
out to be an unpopular idea, and profile-based solutions don't really work 
on the Web.


On Sun, 16 Aug 2009, Henri Sivonen wrote:
> On Aug 15, 2009, at 14:16, Ian Hickson wrote:
> 
> > pubdate="" on <article> is solving the problem that there's no other 
> > way to associate a publication date with a blog entry in HTML, in 
> > particular to allow for conversion to Atom.
> 
> What about making the <article> an item and using a pre-defined itemprop 
> on the <time>?

There's no relationship between the <article> and the fact that it happens 
to contain an item, semantically.


> About the use case itself: Is there implementor interest in implementing 
> the consumption of HTML documents that aren't under the control of the 
> implementor as feeds by mapping the documents to Atom semantics per the 
> rules in the spec?
> 
> (I, for one, have some severely legacy software that runs as a cron job 
> and generated by blog feed from HTML files, but the process requires no 
> spec, since the software is sensitive to conventions I enforce on 
> myself, so there's no need for interop with anyone else's conventions.)

There certainly has been interest in making HTML be self-hosting feeds. 
How much interest there is, and whether it's worth keeping an attribute 
(on <article> or <time>) remains to be seen, I think.

-- 
Ian Hickson               U+1047E                )\._.,--....,'``.    fL
http://ln.hixie.ch/       U+263A                /,   _.. \   _\  ;`._ ,.
Things that are impossible just take longer.   `._.-(,_..'--(,_..'`-.;.'
Received on Tuesday, 25 August 2009 00:55:40 UTC

This archive was generated by hypermail 2.3.1 : Friday, 10 October 2014 16:24:51 UTC