Documet ids (was Archie, WWW access directly to files, and document ids

Tim Berners-Lee (timbl)
Tue, 18 Feb 92 16:44:35 GMT+0100

Date: Tue, 18 Feb 92 16:44:35 GMT+0100
From: timbl (Tim Berners-Lee)
Message-Id: <9202181544.AA20333@ >
To: Brewster Kahle <>
Subject: Documet ids (was Archie, WWW access directly to files, and document ids

From: Brewster Kahle <>

   Date: Tue, 18 Feb 92 14:25:04 GMT+0100
   From: (Tim Berners-Lee)

  ... This relies on the headline [of an archie index] 

	starting with host:/filename.


  I wish we used /filename@hostname so that waisretrieve could handle it (and
  be compatible with the WAIS doc-id.


Well, Brester, I wish that we used //hostname/filename so that it would be
directly compatible with the W3 doc-id ;-). In fact, of course, the user never ses  
the doc-ids themselves as he browses.

But seriously, Brewster, you suggested to John Curran <> that,  
on the subject of document ids, "[Brewster's] proposal that is on the table is  
worth implementing for a good run".  I would suggest that John look at w3 Universla  
Document Identifiers as a similar but more open and more established scheme.  W3  
has been running now for 18 months or so using the UDI syntax and the addressing  
syntax has expanded easily to include wais and now gopher. The www retrieval engine  
will handle any of these, FTP access and news access etc.

When x500 document naming becomes practical, it will be important that the UDI  
scheme can expand to accept x500 names.

There is nothing proprietory or w3-specific about w3 UDIs.  They are generic,  
caonnonical and univeral. Could I strongly suggest that you extend waisretrieve to  
use UDIs?

[Your objection to the w3 UDIs was that you prefered "@" to "//" because you wanted  
to use parsers written for mail.  Is that is a strong enough reason for inventing a  
new, wais-specific scheme instead of using an existing, open one?  Actually, the w3  
scheme uses @ for  //user@hostname/ when a username is needed, which is even more  
mail-like. The choice of punctuation is of course fairly arbitrary.]

The universal document identifier syntax is dead simple. It is described in BNF in  
/pub/www/doc/udi.txt.  Comments from all parties who haven't seen it before are  

We must to put systems together to make the information universe happen, and not  
quarrel about trivia. We must remain open to the future.

I must put down the reasons for UDIs in a paper, but they're probably obvious to  
everyone on these lists....