W3C home > Mailing lists > Public > www-html@w3.org > January 2005

Re: rel="nofollow" attribute

From: Jukka K. Korpela <jkorpela@cs.tut.fi>
Date: Fri, 21 Jan 2005 15:18:03 +0200 (EET)
To: www-html@w3.org
Message-ID: <Pine.GSO.4.58.0501211456580.11033@korppi.cs.tut.fi>

On Fri, 21 Jan 2005, Asbjørn Ulsberg wrote:

> No, but the word «nofollow» is wrong in so many ways that I don't even
> know where to start.

That's true, but the currently "standard" rel values are largely obscure.
And mostly unimplemented. The field is open to players who just love tag
soup, or should we say attribute spices. We can't really blame the search
engines &Co. for putting the rel attribute into prodecural use, when
there's not much logical use for it

> Solutions are good, but someone should have used at least five minutes to
> read the specification text of the 'rel' attribute

Would that have helped? The basic description takes just a few _seconds_
to read, but when you get into the list of actual values, five hours won't
be enough to find what they might really mean and how they could be

> Yes, it's great. But the word still sucks.

It's not a word, it's a string, a "name" in the technical sense. The Web
technologies are full of names that are outright misleading. But besides
that, the word is just great in describing what it _means_. Don't blame
the name for being honest. It is _meant_ to be an instruction to indexing
robots, nothing more, nothing less.

Whether it's great is debatable. This looks like a perfect case of solving
the wrong problem to me. If you don't want links on your pages to count in
indexing and ranking at all, don't put them there. And if you set up a
blog and let everyone and his dog write there, without filtering, then you
_are_ indirectly endorsing any resources they might mention. Not strongly,
but anyway. If they advertize a book, your blog advertizes a book.
Ditto for URLs, whether you turn them into links or not.

What might be logically sound _and_ useful even for the purposes discussed
is an attribute that specifies the relevance of a link in the author's
opinion. It could have a symbolic or a numeric value, though it would need
to get ultimately quantified somehow. You could even set it to zero.
Or you could, as an author, assign different weights to different links.
(Search engines would presumably treat them as relatives weights only,
inside a page.) For example, if your page has a "suggested reading"
section, you could assign different weights to them according to your
judgement (and more weight than any casual links on the page).

But it's probably too late now.

> A better value could have been e.g. «no-credit» or «none» which says that
> the current document does not credit the referenced one («no-credit») or
> does not have any relation to the referenced document at all («none»).

Having no relation at all means it's a link that isn't a link, since a
link is by definition a relation.

But no-credit comes close to my idea of relevance or weight - just the
simplest possible way of defining such an attribute. But it would not be
practical to make it binary. And it is slightly illogical to make an
attribute negative in meaning (reflected in its name), though we have the
precedent of nowrap.

> «nofollow» is just wrong.

Except as name for a procedural attribute, in which role it is just
partly misleading. It doesn't suggest that _users_ shouldn't
follow the link.

Jukka "Yucca" Korpela, http://www.cs.tut.fi/~jkorpela/
Received on Friday, 21 January 2005 13:18:07 UTC

This archive was generated by hypermail 2.4.0 : Thursday, 30 April 2020 16:20:55 UTC