W3C home > Mailing lists > Public > www-html@w3.org > August 1998

Re: LINK Element Confusion

From: John T. Whelan <whelan@physics.utah.edu>
Date: Sun, 16 Aug 1998 12:22:35 -0600
Message-Id: <199808161822.MAA29741@einstein.physics.utah.edu>
To: www-html@w3.org
Odd # of indents = Frank Boumphrey
Even # of indents = Me

	I'm sure Frank knows this, but I didn't: the list of LINK
types defined in HTML 4.0 is in the spec at
<http://www.w3.org/TR/REC-html40/types.html#type-links>; the types
listed in parentheses.  (A good convention would seem to be to use the
primary names listed rather than their synonyms, so START and not
BEGIN.)  The spec says that if you use any others, you should use the
PROFILE attribute of the HEAD element to point to a URI where they're
defined.  However, as noted by the Web Design Group
<http://www.htmlhelp.com/reference/html40/head/head.html>, there is no
format specified for this profile.  Does anyone know of a format
that's in use?  I haven't found any examples of pages that define a
PROFILE attribute (including the WDG page, which uses LINK REV="made",
which is not on the list), besides which most of the work going on
along these lines is addressed at standardizing the META element.

>Actually I think START is just a message to a 'bot' to tell it where to
>begin searching.
> I think it is better to think of "collections of pages" each with its own
>On my private web site, I have a collection of 'font pages', of 'typography
>pages', of 'CSS pages' etc, each with their own index page.
>I should have a minimum of two link tags, one REL=START telling a Bot where
>to start, which would logically point to the root page of the whole document
>collection, and another REL=INDEX indicating the index page of the
>individual collection.

	Which is sort of using INDEX as a substitute for the
non-existent PARENT.  But not every page has an index associated with
it, even stretching the definition of INDEX to mean any page that has
a bunch of links to documents "below" it.  Consider a page consisting
of a set of footnotes for another page.  It's only linked to (in terms
of A elements) by the page for which it provides footnotes (so it
can't have the same INDEX page, although it could have the same
START), but it would be inappropriate to call its parent an index.
Instead, it should have a LINK REV="notes" pointing to the parent
page.  (Except--argh!--there is no NOTES or FOOTNOTES link type

	You're also implictly defining the START page to be where
you'd get if you followed all of the index links on up.  Does that
mean that the page pointed to by REL="start" should not have a LINK
REL="index", since in that case the true start page would be the
target of *that* LINK?  That doesn't seem practical, especially since
I have one index page which lists the various pages I maintain, each
of which is a START page in its own right.  (In a way, you're also
advocating using START as a substitute for a non-existent HOME.)

>> And what about joint endeavors?  It it kosher to have two
>>different LINK REL="start" tags if there are two independent pages
>>which can claim to be the immediate parents of a page?

>I didn't think in OO theory there could be two parents. Each object is a
>"Virgin birth"<g>

	Living in Utah, I feel I should make a joke about polygamy
here, but I'll refrain, as I'm not a native.

>But of course many sites have a neuronal structure, with multiple
>interlinked pages (although I believe this is bad design, but that is
>another story) in which case all pages could claim to be the start page of
>all the others. Probably best to arbitarilly define one page as a start page
>to ensure the poor bot doesn't have a nervous break down!

	There's usually one logical entry, but on occasion there may
be two.  If there are, for example, imagemap and text-based home
pages, or frames and non-frames ones, you could use LINK REL="start"
to point to the main one and LINK REL="alternate start" to point to
the other(s).  But if there are two independent starting points,
presumably a bot would not get too confused if there were two START
LISTs, especially if they had different TITLEs.

	Or another alternative would be to use the BOOKMARK link type
with appropriate TITLEs, so in the case I'm thinking of:
	<LINK REL="bookmark" TITLE="The Big Red What?"
	<LINK REL="bookmark" TITLE="Joe Schlobotnik's Sports Machine"

>As a Professor of relativity of course YOU should have no difficulty with
>neuronal structures<grin>

	Not Professor yet, just Doctor.  Anyway, neural nets are for
ex-physicists. ;-)

>>PS--You left out LINK REV=made, which has been used by at least one
>>browser for a long time.

>FMI: -Which one?

	Lynx, since the early days (despite my complaints, U of U
Physics is still running 2.3 BETA from 1994), has used the HREF of LINK
REV="made" as the Author on the information page ('=' keystroke
command), and also as the URL for comments with the 'c' keystroke
command.  This URL is typically a mailto, but it could also be the
author's homepage.  (So in that case, there could even be a use for
LINK REL="made" if someone wanted to claim authorship of a page for
which they could not set a LINK REV="made".)  (Of course, MADE is also
missing from the HTML 4.0 list of LINK types.)

	Along those lines, do I surmise correctly that there's no need
to label both ends of a relationship with REL and REV LINKs?  And that
the LINK is traditionally put on the end that minimizes the number of
LINKS in any given document.  (E.g., you don't want to have fifteen
LINK REV="index" links in the index document when you could just put
one LINK REL="index" in each of the indexed documents.)

	Getting back to Lynx, version 2.7 does use some of the other
LINK types (e.g., GLOSSARY and COPYRIGHT) to give the user a menu at
the top of the page.
					John T. Whelan
Received on Sunday, 16 August 1998 14:22:13 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Tuesday, 27 March 2012 18:15:37 GMT