Re: Persistent caches - was: Are *relative* URIs as namespace nemes considered harmful?

Tim Berners-Lee wrote:

> (When the
> Internet was very young, my best beloved, then there were not very many
> nodes on it.
> They had names and numbers and unix machines had a file called /etc/hosts
> which contained a list to convert from names to addresses. The problem of
> maintaining this file led to the creation of the  Domain Name System.  Now
> you can relive the early days with  catalog files to map between the names
> and addresses of your favorite things).

Foo.  All that the DNS does is provide a bunch of /etc/hosts servers
with machinery for delegating requests from one to another.  (Yes, I'm
oversimplifying for effect; so are you.)

FPIs now allow delegation mechanisms, so nothing prevents us from setting
up a bunch of FPI root servers that know how to delegate down the FPI
tree.   All that's really needed is a simple server-to-server protocol,
a well-known port, and some commitment.  Just what DNS has, in fact.

> If the social conditions (control, persistence) are things
> this community really likes, then it would be quite simple to define a URI
> scheme fpi: which would fix this.  That could easily become a W3C activity
> - so long as they could be justified as giving something which existing
> URI schemes don't.

What they have (and I know the mechanism has been broken for a long time,
and still is, but that *can* be fixed) is a registrar that's committed:

	1) to hand out names on request (this is the broken part) and
	2) not to hand out the same FPI-prefix twice (this is the good part).

The DNS registrars are committed to
just the opposite: if you don't pay your US$50 a year or whatever, they
can and will take your domain name away and assign it to someone else.
The domain 'spam.org', for example, has had four owners that I know of.

This makes complete hash of any kind of naming scheme, since the new owners
inherit no sort of commitment to keep the old names in being.  Imagine the
bibliographic chaos that would result if publishers had to pay "maintenance
fees" to keep their ISBN prefixes, and if they didn't, the ISBN authority
would happily reassign those prefixes to some other publisher!
You could never again depend solely on the ISBN of a book to identify it,
but would have to know as of what date the reference was made, and refer
to a historical database mapping ISBN prefixes to publishers.

Of course this doesn't happen, since there are no "desirable ISBN prefixes"
that people are willing to pay big $$$$ to gain, or go to court to
confiscate.  But that does happen with domain names and will continue
to happen with increasing frequencies.  When the URL was invented,
DNS names were still once-and-for-all: no longer.

With ISO-registered FPIs, we have *one more chance* to get it right this
time.

> >If names are intended, then the string
> >value of an absolute uri is sufficient (ignoring the uniqueness debate).

Only if augmented by a date, as explained above.

> Making a generic API for this would be a useful technique.
> (An XML file format for the persistent cache directory -- catalog --
> would of course be useful).

(*cough*) See http://www.ccil.org/~cowan/XML/XCatalog.html

-- 

Schlingt dreifach einen Kreis um dies! || John Cowan <jcowan@reutershealth.com>
Schliesst euer Aug vor heiliger Schau,  || http://www.reutershealth.com
Denn er genoss vom Honig-Tau,           || http://www.ccil.org/~cowan
Und trank die Milch vom Paradies.            -- Coleridge (tr. Politzer)

Received on Friday, 19 May 2000 10:24:07 UTC