Mark Day: How was the process or convention used for naming web servers originally developed?

Roy Fielding: There were two reasons for this.  First, the Web server at CERN was called info.cern.  There were a couple of different protocols there including Telnet, FTP, etc.  They wanted to split the services over several machines.

Second, name consistency was important. They wanted the capacity for moving the website throughout the organization, but also maintain the same address. For example, at UCI there are different names for the WWW and FTP server, www.ics.uci.edu and ftp.ics.uci.edu yet they point to the same machine.

In 1993 NCSA chose a name for the WWW server.  Several names were considered:

www - distinctive - no one else had this name, but it was difficult to pronounce.
Web - pronounceable but not unique
W3 - web project at CERN

NCSA advised people to go with www.  Since NCSA was influential with the most popular web server at the time (HTTP), everyone decided to follow the www convention.

Walt Scacchi:  Why are Port numbers in URLs?

Roy Fielding: The port number was decided by the naming authority for web servers using TCP.  The standard that was decided for web servers was 80.  However, Roy's test server, Kiwi, is at port 8080.  It was the implementation of TCP that required the assignment of a port number.

There were problems with using TCP.  Issues of authority restricted use of web servers.

Walt Scacchi: I understand that, but why the need for a port numbering scheme?

Roy Fielding: It's because of privilege issue and not the issue of having multiple servers.  Since Roy didn't have sufficient privileges to his machine, he had to set up the server on a high number port.  MIT students (who published the more interesting web pages) had the same issues and conflicted with their support personnel.

Daniel LaLiberte: How about alternatives to URL's to URNs and scaling issues?

Roy Fielding:  I left a lot of stuff out from my talk because I know Larry Masinter and Michael Mealing are coming after me.

Daniel LaLiberte:  How about the history?

Roy Fielding: The initial proposal by Tim Berners-Lee is a listing of what was implemented for the web.  Things that are names and locations.  Names continue to be valid identifier, despite location.  Name is consistent to the resource that you're referring to.

As various proposal brought to IETF, the most appropriate resource. Spun off URI in 1993, and that working group decided their should be a distinction between names and location and syntax between the two.

The only thing that makes things persistent is the naming authority. Someone who will need to last throughout eternity.  5 years is
typical.  Organization that maintains these identifier - expensive task is when the URI become fruitful

Scaling Issues - Larry Masinter will talk about that.

Distinguishing feature name tend to be validator

Various proposals:

IETF - identification types most appropriate URI working group formed in 1993 and that working group should be a distinction.  Different Syntax for names and identifiers

Eternity - identifiers being persistent, 5 years for Identifier persistence is a very expensive task.

Put off syntax - valid if name or location same syntax

Bob Morgan:  U is for Universal vs. uniform? do they mean the same thing?

Roy Fielding: Way back, it was originally named "universal" by Tim Berners-Lee.  But he said that uniform means what he meant by universal.

Larry Masinter: Was intended to be uniform. In the IETF, the "U" is uniform.

Unique vs. universal
Everything in the IETF is uniform.

Rohit Khare:  URCs were throwing everything into a bag, but XML has taken over.

{ed: the notes become bit confused at this point}
History?
Listing of names
Tim Berners-Lee
Universal vs. uniform
Tml should be uniform Uri working group product
2936 - should be uniform