HTML link "Made" relationship

Tim Berners-Lee <timbl@www3.cern.ch>
Date: Tue, 11 May 93 16:46:44 +0100
From: Tim Berners-Lee <timbl@www3.cern.ch>
Message-id: <9305111546.AA03707@www3.cern.ch>
To: grobe@ukanaix.cc.ukans.edu (Michael Grobe)
Subject: HTML link "Made" relationship
Cc: www-talk@nxoc01.cern.ch
Reply-To: timbl@nxoc01.cern.ch

> From: grobe@ukanaix.cc.ukans.edu (Michael Grobe)
> To: www-talk@nxoc01.cern.ch
> Date: Mon, 10 May 93 14:49:41 CDT
 

> chapter 6 of the march 15, 1993 version of "Hypertext Markup  
Language
> (HTML)" by Berners-Lee and Connolly discusses "Link Relationship  
values."
> 

> i have several questions about this chapter:

> first, what is its status? are there browsers in use that recognize
> these link attribute values? have any of these attributes been  
promoted
> to be part of the standard?  are they likely to become part of the
> standard? if so, when and under what authority?

They have not been promoted to part of the standard, but if they
are tested they could be.  RFCs don't become standards without
implementations.  No browsers that I know of implement them though
many people have expressed great enthusiasm.  The thing will go
in RFC direction, the authority will therefore be IANA and the IETF
or whatever officially is the authority in the nwo (IAB? ISoc?)

> second, if browsers ignore attribute values that they do not  
recognize, why
> must these "experimental" values be preceded with "X-".

This scheme (which is used for many Internet standards) means
that ifyou invent a scheme which puts a certain semantics on
a new relationship, you can guarrantee that the scheme won't break
because someone just happens to use it for something else with
different semantics.  During the test phase, you can't be SURE
that noone else thought of the same X-relation, and that their
documents will totally confuse your software.  (though it is
unlikely especially with a this list running)

> third, are there any example documents showing the use of these  
values?


Nope (correct me?).  [[Historical note:  The first hypertext system I  
wrote used typed links.  An interesting result was
that though I had initially imagined that the number of link
types would grow without bound, in fact I settled down with a
dozen or so having all the semantics there seemed to be
beteween the things I was making notes about]].

> fouth, is the value "Made" correctly described in this chapter?
> the chapter says that Made means that the:
> 

>    Person (etc) described by node A is author of, or is responsible  
for B
> 

> where A is the source document and B is the destination document.  
however, 

> the chapter also says that one use for the value is "for sending
> mail to authors," and it seems to me that Made would be more useful  
if
> A and B were reversed in the definition above.
> that is, it seems more useful for a document to contain a pointer  
to a file
> describing an author where the link relationship would be "Made"  
with
> the understanding that the person described in B is the author of  
A.
> can someone clarify this (what seems to me to be a) discrepancy?

Every link type can be used in a REL attribute or in a REV
attribute (or in some -- non acyclic -- cases conceivably both).
The REV attribute is just the same, but the direction is
reversed.  This saves having to have a convention for
naming reverse link types such as "made-by".
 

> fifth, since we are looking for a way to connect html documents  
with
> their authors or responsible authorities, the value Made appears to  
be
> a possible approach.  but, i would be interested in knowing any  
other
> conventions for recording such information that may be planned or  
under
> discussion.  i would actually prefer (at least intuitively) to have
> a tag defined for the document <head> section that could be used to
> record responsibility for the document.  for example, <owner> could
> be defined and have name, organization, and e-mail address as  
attributes.
> our browser could then pick up the owner e-mail address for posting
> comments from users reading a particular file directly to the  
owner.
> in particular, we might have the following tag 

>      <owner name="Michael Grobe" e-mail="grobe@kuhub.cc.ukans.edu">
> comments?


Whilest any useful additions to HTML for HMML are open to
discussion now, I would like to see more generic forms used.
This is because HTML is not an application-specific document
type, it is in fact a great big compromise which allows enough
fairly generic structures to represent most things anyone wants.
If "owner" can be represented as a generic relationship, between
a document and a person, with a particular type, then I would
prefer the generic form to be used rather than a specific tag  
invented.  I think this will make processing easier, also expansion,
and keep the DTD small.

<LINK REV=MADE HREF="whois:S=Grobe;G=Michael;O=Ukans">

Here is one example of something which is
easier to do is a web traversal filtered by relationships. Consider
"Give me a list of (and fire :-) all the people who have MADE  
software which is USED BY this program and for which there is no  
document which DESCRIBES that software" 

or, "Print out all documents which this document recursively INCLUDES
to any depth", and so on.

> :michael grobe
> academic computing services
> the university of kansas
 


Tim Berners-Lee
CERN