 |
 |
> Date: Wed, 15 May 2002 11:04:31 -0400 > From: "LeVan,Ralph" <[log in to unmask]> > > > My proposal is the following: > > > > 1) Allow local and global namespace prefixes. > > 2) Let the server decide what to do with unprefixed index names. > > 3) Let's try to define a list of unprefixed index names > > (instead of dc.xxxx) just to standardize the names (ambiguity > > is the clients risc). > > 4) Make sure Explain specifies the servers behaviour unambiguously > > 1, 2 and 4 are great! > > I'm not at all happy with #3. That's just Bib-1 all over again. > The ZIG has no business defining access points for other > communities. I think Theo's point is that we're talking about access points that don't belong to any community -- or, if you prefer, the "no community" community. I agree with him that we should provide a lax way for talking about things things (and add that there can be no strict way, for fundamental reasons.) > Date: Wed, 15 May 2002 17:24:14 +0200 > From: Theo van Veen <[log in to unmask]> > > Ok, we are almost there. > > For #3: > For unprefixed names let's use as much as possible the element and > qualifier names from the available DCMI application profiles without > the namespace prefix (ambiguity is the clients risc). I think this is a neat compromise. To state it more clearly (if I may): when a server receives a search against an unqualified index _for which is has no semantics of its own_, it is gently encouraged to treat it as semantically similar to the same-named Dublic Core element. (Insert additional obfuscatory prose according to taste.) Next please! > Date: Wed, 15 May 2002 16:01:42 +0100 > From: Robert Sanderson <[log in to unmask]> > > Here's my list of names. > > <#include "bib1.h"> (many hundred) > <#include "collectable_card_games.h"> (many thousands) > <#include "email.h"> (~100) > <#include "zthes.h"> ? > <#include "network.h"> (~10) > <#include "tei.h"> (many many thousands) > <#include "archives.h"> (see tei) > <#include "CIMI.h"> (see tei) > <#include "artworld.h"> (see tei) > <#include "OMRAS.h"> (see tei) > <#include "lego.h"> (see tei) > <#include "userregistry.h"> (~100) Way, way too many. We couldn't possibly require -- or even encourage -- servers to standardise semantics of that many access points. I was think more in terms of, ooh, let's say, about fifteen. OK, nearly done ... > Date: Wed, 15 May 2002 16:31:41 +0100 > From: Robert Sanderson <[log in to unmask]> > > What's the Dublin Core for collectable card rarity? There isn't one. > Or the name of the card set that it's from? There isn't one. > How about the DC for username? port number? Music tempo? How do > you specify the name of the journal that an article is to be found > in? (I'm going to leave you to figure out the answers to these :-) > Why is DC to be treated as superior to any other indexset? Because it's there. What we're talking about here is a set of semi-standardised interpretations for access points which can be used in cross-domain-like searches. Seems daft to make some up when we already have a perfectly sensible set lying around. > If I have a default indexset in a collectable card database, then > searching for unprefixed 'set' should not have to be interpreted as > searching for a mathematical set or anything else. No indeed -- servers would say: "Sorry, mate, got no idea what you're talking about." > I still disagree fundamentally. BIB1 proves that This Does Not > Work. It proves that BIB1 doesn't work. That's all. _/|_ _______________________________________________________________ /o ) \/ Mike Taylor <[log in to unmask]> www.miketaylor.org.uk )_v__/\ "Whole AND segments" -- Monty Python.