Comments on Hypertext Links in HTML

Jukka Korpela,

These comments reflect the author's personal views. They are related to W3C Working Draft 28-Mar-97 Hypertext Links in HTML.

These comments suggest, in addition to other improvements, standardization of REV and REL values within HTML specifications as opposite to leaving them up to "profiles".


The link concept, in spite of its being so crucial in hypertext systems, is still very primitive in HTML. The anchor (A) element gives us the basic idea of some construct in some document pointing to another document or to a location in a document. The semantic relation and the point of the pointing remains unspecified.

Consider, for example, a link with link text Telnet in the following context:

We recommend that you use the Telnet program to connect to our computers.
Looking just at the above text as rendered by a Web browser, what are the odds for the user guessing correctly what happens if he follows the link? Different things might happen; for instance any of the following might be a sensible in some context:

An educated user might look at the URL corresponding to the link text, since that URL is typically shown by a (graphical) Web browser, or he might specifically request the browser to show it, or even look at the HTML source code. But that would be technical information only, telling whether we have an http URL (eg to an .html file, probably containing HTML, or to a .txt file, probably containing plain text, or to a .doc file, possibly containing data in a text editor format), or an ftp URL (pointing to a server, or a directory, or a file) or a telnet URL, or something else. The real point of pointing may still remain unknown, since the scheme or other components or the URL cannot tell how the linked resource relates to the current one, although they may sometimes give some clue about it.

An anchor (A) element can contain a TITLE attribute which provides an "advisory title for the linked resource". It is, however rarely used and not widely supported by browsers. In addition to that, it could at best be a useful auxiliary note, not a basic designator of the nature of the link. Being arbitrary plain text written by the author, it cannot serve the purposes of automatically detecting the relationship between the linked resources or the nature of the link. And users should not be expected to be able to decipher each author's personal way of telling in hiw own words the nature of the link.

There have been various attempts to improve the situation. The draft we are discussing here adds little to existing HTML specifications in this respect. It suggests that REL and REV values, used to specify the semantic relationships between linked documents, be defined in "profiles". Thus, HTML would just contain a method for specifying where is the profile which defines the REL and REV values. This looks like final surrender: no attempt would be made to standardize even some basic relationships Webwide.

Therefore, these comments of mine concentrate on

There are also some less important comments on other issues in the draft.

Why should we standardize REL and REV values?

A draft specification of REL and REV values

A draft style guide: How to use REL and REV

Other comments

Jukka Korpela
June 12th, 1997