Re: Trees have one root

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



At 09:09 PM 7/29/2002 -0700, Peter Deutsch wrote:
>I believe it's time the
>IETF recognizes that it's time to move beyond what are
>really the politics of ICANN and focus on the technical
>issues surrounding extending the current technologies to
>make them more useful. If we do that without touching the
>legacy systems for now, fine but this would imply the need
>to set up some alternative roots for experimentation and
>proof of concept. 

peter,

I would suggest that exploration need not require an
"alternate root" if the extension is within the framework
of DNS (new class of object for example). John Klensin
and others have expressed themselves on the merits and
weaknesses of that approach. What is important, I think,
is to avoid the creation of conflicting naming schemes
within a given framework. However, if we begin by asking
the question, "what objects need identifiers that we do
not or cannot now identify?" we might indeed discover 
alternative identifier mechanisms. It would be helpful
for a variety of operational and probably technical reasons
for any such new scheme not to conflict syntactically with
existing ones, I believe. 

The Directory-service notion is one dimension in which
to explore. Keywords are another although thus far they
have understandably required a context in which to be
interpreted since they all appear to have a common and
simply syntax (cf: AOL keywords, Real-Names keywords,
etc). 

The DNS provides a method for associating a relatively
simple syntactic construct with a host on the Internet
(or an internet). These names can be embedded within
the larger construct of URLs to refer to objects and
processes that exist on (in?) those hosts. 

One might want a richer vocabulary for referencing 
objects (cf. Bob Kahn's Handle System) in which the
identifier is associated not only with an object but
also with a variable amount of meta-data about the
object's ownership, accessibility, etc. 

It strikes me as a useful principle to try not to
make things like DNS more complex in seeking to explore
and invent these new concepts if only to avoid making
existing and relatively simple things more complex. 
Domain names seem to work pretty well for hosts and
WWW URLs, but in the latter case, a broken link caused
by the disappearance of an object at a host or even the
host itself makes the longevity of the binding less than
satisfactory. Perhaps other methods of expressing and
implementing the binding of name to object would be
useful, but I would argue against making DNS more complex
to accommodate, if this extension also made the pre-existing
use of DNS more complicated. 

The experience with the notion of Internationalized Domain
Names has made clear how dependent DNS is on the simplicity
of the USASCII encoding and how hard it is to make what one
might have thought would be a simple extension to a richer
range of character encodings. 

So to reiterate, I believe it would be a very useful exercise
to catalog the various name/thing binding notions and to try
to articulate what additional functionality we might wish to
achieve, then examine options for achieving them without 
breaking or making unnecessarily more complex our existing
binding tools. That view suggests that "alternate root" may
not be the right starting point, if by that term one means
alternate root of the same DNS mechanism we already have. 
The syntactic and potential semantic confusion (e.g. using
an identifier in the wrong context) does not seem worth the
price. However, a new identifier space whose context can be
clearly placed through syntactic clues, might provide a 
platform for exploration without or at least with less disruption.

Vint



[Index of Archives]     [IETF Annoucements]     [IETF]     [IP Storage]     [Yosemite News]     [Linux SCTP]     [Linux Newbies]     [Fedora Users]