W3C home > Mailing lists > Public > www-html@w3.org > November 2001

Re: Semantics of <link rel="next">

From: fantasai <fantasai@escape.com>
Date: Fri, 16 Nov 2001 23:35:36 -0500
Message-ID: <3BF5E918.EA80B083@escape.com>
To: Ben Bucksch <ben.bucksch.news@beonex.com>
CC: www-html@w3.org
Ben Bucksch wrote:
>
> <http://www.w3.org/TR/REC-html40/types.html#type-links>
> 
> > *Next*
> > Refers to the next document in a linear sequence of documents.
> >
> Unclear.

Actually, I think this is the best definition--it's both clear and concise.
A 'next'-type link has two requirments:
  1.) a *linear sequence* of documents
  2.) links to the next document in that sequence

> <http://www.ics.uci.edu/~ejw/authoring/draft-ietf-html-relrev-00.txt>
>
>   3.1. Siblings
> 
> This method is most useful for sites/documents, which take the hierarchy
> (and hyperlinking) to an extreme: Going deeper in the hierarchy is
> optional and only useful, if more in-depth information to a certain
> topic (page) is wanted by the reader. The parent is a summary of its childs.
> 
> In such a site, it makes completely sense to go to the next sibling
> without reading the childs. Reading a certain child or all childs would
> be an active decision of the reader, not the default one. Thus, Next
> would point to the next sibling, not the first child.

Yes. However, this only applies where there is a sequence in the siblings--
if there is no order inherent in the siblings, then you should not use
'next' to navigate among them.

> Now, if I really flatten the whole site, creating a linear tour through
> the whole site and using that string as the "set of pages", I end up
> with a First that is always identical to Start and a Last that always
> points to the same random page that happens to be last in my string. In
> other words, First and Last would be useless.

Yes, and this is a case where you don't have a sequence, and thus shouldn't
use sequential link types.

> Note that while Next here creates a tour, it is not necessarily a
> *guided* tour. If you do it most ________logical________ (generated),
> the tour goes through the whole site, no matter how uninteresting the
> pages are. //emphasis mine

A logical set of links would also account for the content and thus use an
appropriate link scheme for that content. The word you seek here is, I
believe, "systematic".

> OTOH, this interpretation makes a lot of sense for documents (broken up
> into several HTML pages) with a structure of traditional books, which
> are read linearily.

Agreed.

> However, this approach has the severe disadvantage that it creates a lot
> of inconsistency. It is hard to predict for the visitor, if he should
> trust the author to create a nice guided tour, which he should follow,
> or if the author generated the link, with variing quality as the result.

Guideline: Only use 'next' <link>s if you'd have otherwise put a 
           <a href="..">Next</a> link.

> Since all approaches make some sense, my suggestion is to support all of
> them explicitly in the markup. I suggest
> 
>    1. "sibling next"
>    2. "linear next"
>    3. "guided next"
>    4. "custom next"

'rel' and 'rev' take a space-separated _list_ of link types; 'guided'
isn't much of a link type.

Also, this is, IMO, overdefining the 'next' link type. Not everything
should be defined with absolute precision---some things are better left
open to interpretation. This allows for flexibility and adaptability.

I'd leave the exact usage of a 'next' link at the author's discretion.
It's defined enough to be unambiguous to the audience (your average user
won't care whether it's "linear next" or "guided next" or "custom next"),
and loose enough to accommodate a variety of situations.

(moved)
> The [IETF draft] is the only one (later) mentioning hierarchies at all.
> It specifies "sibling" for relationships as in 1 [Does Next go to the
> next sibling (node/page with the same parent)?]. There's an interesting
> example:
> 
> > <A REL="SIBLING NEXT" HREF="..." > next </A>
> 
> This sounds logical, but shifts the problem somewhat to the UA. Current
> implementations behave unknown (and sometimes odd), if there is both a
> "next" and a "sibling next" rel.

My comments on Chris Hoess's draft for Mozilla included some notes on
sequences and hierarchies wrt. link types. You might find them
interesting:

  fantasai. "Re: Draft spec for Link Types", netscape.public.mozilla.ui
      (2001-10-17). message-id: <3BCD6ABB.5C7CE8B4@escape.com>
  http://groups.google.com/groups?hl=en&selm=3BCD6ABB.5C7CE8B4%40escape.com
   (note - line 112: "first next parent" should read "first prev parent".)
Received on Friday, 16 November 2001 23:33:08 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Tuesday, 27 March 2012 18:15:49 GMT