Re: Symbolic vs Numeric identifiers (was Re: URL internationalization!)

Gregory J. Woodhouse (gjw@wnetc.com)
Mon, 24 Feb 1997 17:14:14 -0800 (PST)


Date: Mon, 24 Feb 1997 17:14:14 -0800 (PST)
From: "Gregory J. Woodhouse" <gjw@wnetc.com>
To: Daniel LaLiberte <liberte@ncsa.uiuc.edu>
Cc: uri@bunyip.com
Subject: Re: Symbolic vs Numeric identifiers (was Re: URL internationalization!)
In-Reply-To: <199702241629.KAA04178@void.ncsa.uiuc.edu>
Message-Id: <Pine.SGI.3.95.970224165322.7943A-100000@shellx.best.com>

On Mon, 24 Feb 1997, Daniel LaLiberte wrote:

> Gregory J. Woodhouse writes:
>  > I'm not so sure they are. True, existing applications use URLs like
>  > filenames, but then again, that's all we have. If there were no DNS, we
>  > would use IP addresses [...]
> 
> Another way to describe this division is symbolic versus numeric
> identifiers.  URLs are already symbolic, and as such, they belong in
> the column with filenames and domain names.
> 

Actually, I think the fact that URLs are symbolic is quite incidedntal to
this argument. The fact is, a URL like

http://wherever.org/whtever.html

is symbolic merely reflects the fact that the referenced web page is
fetched by connecting to wherever.org on port 80 and sening a command like

GET /whatever.html HTTP/1.1

In other words, the URL is symbolic because the protocol elementnts like
the hostname and the request-URI are already symbolic. What i-nodes and
URLs have in common is tht they both contain the information actually used
in retrieving a file. If you will, they are both *locations* and the fact
tht one is symbolic in form is simply an artificat of the protocol(s) used.

> Curiously, some people want URNs to be numeric to avoid all possible
> semantic interpretations by humans.  Note that an i-node is the *real*
> reference to a piece of data that may have many different file names,
> and an IP address is the *real* identifier of a machine that may have
> many different domain names.  Now a URN is the *real* identifier of a
> resource that may have many different URLs.  But then shouldn't URLs
> be mapped to URNs instead of the other way around?
>

I can sympathize with this desire, but I have yet to be convinced. The
problem is that the URN of a resource may be the "real" name is not sense
that it is the official name, but an i-node is a "real" refernce to a file
in the sense that it gives the physical location of the file. These senses
of "real" are quite different. 
 
> People who want numeric URNs also envision a higher layer of human
> friendly identifiers that would be mapped to the lower level numeric
> URNs.  Where do URLs fit in this scheme then?  There are lower layers
> of identifiers under i-nodes and IP addresses.
> 

As I've indicated above, I think this is exactly backwards. The URL encodes
protocol specific information on how to actually retrieve a resource. True,
a single resource can have multiple URLs (for example, in the case of hosts
with multiple IP addresses, or HTTP/1.1 servers which use Host to implement
virtual domains), but this doesn't change the basic fact that describe how
resources are accessed via the appropriate protocol. URNs, on the other
hand resolve to URLs, just as directory entries (where I mean Unix
directory, not X.500) reference i-nodes.

> I should point out that it is not necessarily the case that "symbolic"
> is to "numeric" as "name" is to "location", but people like to use
> them that way.  At this point (in this message), I'm not saying which
> way things should be - I would like to see this all brought out into
> the open for discussion since there are some assumptions people are
> making about how things should be done.
> 

Agreed.


> --
> Daniel LaLiberte (liberte@ncsa.uiuc.edu)
> National Center for Supercomputing Applications
> http://union.ncsa.uiuc.edu/~liberte/
> 

---
gjw@wnetc.com    /    http://www.wnetc.com/home.html
If you're going to reinvent the wheel, at least try to come
up with a better one.