Date: Wed, 9 Feb 94 18:26:06 +0100
From: Tim Berners-Lee <timbl@ptpc00.cern.ch>
Message-Id: <9402091726.AA04494@ptpc00.cern.ch>
To: Karen R. Sollins <sollins@lcs.mit.edu>
Subject: Re: URN functional spec
Karen,
Pardon my dumbness, could you clarify the meaning of
"Scalability implies that a scheme for devising names may be scalable both at
its terminators as well as within the structure."
This sounds a bit like what I have been looking for a name
for -- inverse modularity -- or whatever -- fitting into a
bigger whole as well as being well divided into parts. Or
are you talking about an name which is implicitly a
segmented string and some property of the string? Or some
property of a hierarchy? Are saying that the product of a
fixed number of non-scalable spaces does not scale either?
(eg a concatenation of two strings each a flat space)? If
you are saying more, then perhaps it should be more
explicit. You have derived a requirement for a hierarchical
naming authority structure, but you have not deduced a
hierarchical structure for the name. Are you trying to avoid
this? If so, why -- is there a (serious) alternative?
A terminator sounds like character (or Arnold Schwartzenegger(sp?)).
You define scalability as the requirement that "URNs can be assigned
to any resource that might conceivably be available on the network, for
hundreds of years". There seems to be a bit of jump here to terminators.
_____________________________
I expressed a separate requirement on the list a while ago
which I think is important, that if a naming authority
chooses to use a scalable naming scheme, then it may
(and may be encouraged to) define it in such a way that
naming authority can be retrospectively delegated for parts
of the scalable space, with no effect on existing names.
Similarly, a naming authority may wish to subsume parts
which it had previously delegated. This should be flexible.
(In a hierarchical name space expressed in the conventional
way as a segmented string, this means that the delimiters
between the naming authority specifier and the part allocated
by the authority should NOT be distinguishable from
the hierarchical delimiters within either the naming authority specifier or
the part allocated by that authority.) I feel this
is a requirement for the good scaling of the URN. You ask good
scaling of the authorities, you should make sure it is
provided by the URN in the large.
_________________________________________
nit:
You assume that the encoding for presentation for people and for
transmission will be the same. Reasonable, but an unstated assumption.
_________________________________________
Alan had asked for stuff which is in your document to be removed
from the URL document -- I will make reference to yours then
for a definition of URLs and URNs.
Tim