Eliot Christian, leader of the GILS effort, found fault with certain of
my statements concerning the relationship of DC and GILS. His own
responses are included below.
stu
> -----Original Message-----
> From: Eliot Christian [SMTP:[log in to unmask]]
> Sent: Thursday, March 19, 1998 5:12 AM
> To: Weibel,Stu
> Subject: Re: FW: Dublin Core versus GILS
>
> Hi Stu
>
> At 03:53 PM 3/18/98 -0500, you wrote:
> >
> >Eliot,
> >
> >here is the note I sent. I won't be in the least offended if you
> >correct my misstatement with a note to the list.
> >
> >If you would rather send me a note of correction that I would send to
> >the list, I'd be happy to do that as well.
>
> Thanks, I'd appreciate that and will respond below.
>
> >> -----Original Message-----
> >> From: Weibel,Stu [SMTP:[log in to unmask]]
> >> Sent: Tuesday, March 17, 1998 4:11 PM
> >> To: 'James Douglas Watson'; [log in to unmask]
> >> Subject: RE: Dublin Core versus GILS
> >>
> >>
> >>
> >> > -----Original Message-----
> >> > From: James Douglas Watson [SMTP:[log in to unmask]]
> >> > Sent: Tuesday, March 17, 1998 10:08 AM
> >> > To: [log in to unmask]
> >> > Subject: Dublin Core versus GILS
> >> >
> >> > Hello met2 listservers:
> >> > I am working under contract for a Canadian federal department. I
> am
> >> > trying to locate
> >> > information on the pros and cons of the use of Dublin Core vs.
> GILS
> >> > for the creation
> >> > of metadata to be inserted in the department's web site. I have
> >> > looked at the DC
> >> > home page, the DC-5 web site, and the web site on the latest
> project
> >> > by the Australian
> >> > government that is using DC.
> >> > I got some excellent listserv responses from another listserv
> >> dealing
> >> > with Australian
> >> > metadata.
> >> >
> >> > Am I right in thinking that DC can be mapped to MARC and GILS,
> but
> >> > that it is not
> >> > as mappable as GILS.
> >> >
> >> [Weibel,Stu] My understanding of GILS semantics is that they
> >> are largely based on MARC/AACR2, and hence mapping between them
> should
> >> be straightforward. DC semantics have been mapped to MARC by
> Rebecca
> >> Guenther at the Library of Congress, and she is perhaps the best
> >> person to comment on the differences.
>
> Yes, there is a direct mapping of GILS elements to MARC elements.
> Also,
> the GILS Profile provides that the MARC record is designated as being
> generated via GILS. This is important to allow such records to be
> intermixed with a library's other MARC records that follow AACR2 or
> other cataloging rules.
>
> >> > Am I also right in thinking that at the moment DC is not as
> >> > Z39.50 compliant as GILS.
> >> >
> >> [Weibel,Stu] Dublin Core elements have attracted considerable
> >> attention in the Z39.50 community and are being included in the
> BIB-1
> >> attribute set.
>
> I think the straight answer to the question is "yes".
>
> There is continuing discussion in the Z39.50 community about how to
> address the Dublin Core elements. I think you are going to need a
> Dublin Core Profile in Z39.50, not merely the addition of Dublin Core
> elements in Bib-1. For one thing, Bib-1 is not required for Z39.50
> compliance. Also, Bib-1 does not specify what needs to happen if,
> for example, you want a MARC representation of Dublin Core
> elements from Z39.50.
>
> >> > Am I also right in assuming that although GILS is a lot
> >> > more complicated, it is superior in terms of being able to
> describe
> >> > information more fully.
>
> GILS and Dublin Core are different things, so it's a bit artificial to
>
> compare them directly.
>
> Since GILS is a search interface, I would have to compare DC and GILS
> on the basis of what a search server needs to support. A
> GILS-compliant
> server has to support search on 8 elements (Title, Originator,
> Distributor, Subject Terms, Date of Last Modification, Record Source,
> Local Control Number and full-text). I suppose a Dublin Core search
> service would be expected to support search against all 15 of the
> unqualified Dublin Core elements. (I guess at some point there will
> emerge a consensus on additional qualified elements that should be
> supported for searching as well.)
>
> The GILS Profile lists by their registed number about 150 well-known
> metadata elements also available, but it does not require that these
> be searchable. (The numbers are essential to encourage cross searching
>
> mutiple languages and other communities of interest.)
>
> >> [Weibel,Stu] There are potentially many niches in the resource
> >> description ecology. DC and GILS are two such niches, each
> targetted
> >> to slightly different applications, and there is overlap between
> >> them. Eliot Christian, the major champion for GILS, visited a few
> >> months back and one of the things we agreed about is that
> implementing
> >> DC could be thought of as a manageable step towards implementing a
> >> fully-functional GILS server. There is no inherent conflict
> between
> >> them.
> >>
> >> The overall effectiveness of any resource description strategy
> >> is a compound metric made up of considerations of:
> >> - cost of production and maintenance
> >> - effectiveness for discovery (precision and recall)
> >> - degree of interoperability
> >> - accessibility of the congnitive model to users
> >> - I've probably missed some others...?
> >>
> >> I don't think you will easily reach a defensible conclusion on
> >> this multivariate scale concerning the relative merits of GILS and
> >> DC, given that doing so would require far more data than now
> exists.
> >> Given that there is no fundamental incompatibility between the two,
> >> deploying DC as part of a larger GILS oriented strategy would keep
> >> you in sync with both, and would improve prospects for
> >> interoperability with many bleeding edge resource description
> >> projects in at least 10 countries world-wide.
> >>
> >> Please note that I do not assert this interoperability as a
> >> certain consequence of deploying DC. The DC community has a great
> >> deal of work to do to demonstrate operational interoperability
> >> across domains and languages, but I am encouraged that there are
> >> many people in many countries working towards these goals.
>
> Very nicely stated, and I agree completely with what you say.
>
> I would add that GILS begins with the notion that there are already
> many valuable collections of metadata that use all sorts of different
> elements. We do not desire to homogenize these metedata. We assume
> they are maintained by a community of interest for some particular
> needs that are typically more specific than what is needed by
> outsiders.
> In GILS, the original elements are not changed but are merely mapped
> to
> a well-known attribute that has a good match semantically.
>
> It would be a real shame if folks actually replace existing metadata
> with GILS or Dublin Core elements and effectively "dumb it down"!
>
> >> > I am very interested in getting some advice on the nitty gritty
> >> > technical differences between the two (please keeping in mind
> >> > that although I am a professional librarian, I am not a technical
>
> >> > expert?) Also, it seems to me that with the growth of the WWW
> >> > that DC seems to be taking off while my supervisor has been
> >> > hearing that GILS is dying in the U.S. Has anyone heard about
> >> > the prospects for GILS in the U.S.
> >> >
> >> [Weibel,Stu] It is my understanding that the Federal
> >> Information Processing Standard that supports GILS deployment in
> >> the US will expire this year and will not be renewed.
>
> In actual fact, the Federal Information Processing Standard for the
> GILS
> Profile did not expire, it was re-issued this year as FIPS 192-1. (The
>
> re-issuance was to update from GILS Profile version 1 to version 2.
> Copies
> are available from the National Institute of Standards and
> Technology.)
>
> The U.S. Federal GILS continues to be in force under U.S. Public Law
> and
> policy, as was reiterated again to all Federal agencies by the Office
> of
> Management and Budget in a Memorandum to agency heads dated Feb. 6,
> 1998.
> <http://www.usgs.gov/gils/omb98-05.html>.
>
> You should also note that there are many other GILS implementations
> aside
> from the U.S. Federal GILS. Some of these are noted at
> <http://www.usgs.gov/gils/sampler.html> and many of these have a
> formal base such as state law or policy.
>
>
> Eliot Christian, US Geological Survey, 802 National Center, Reston VA
> 20192
> [log in to unmask] Office 703-648-7245 FAX 703-648-7112 Home
> 703-476-6134
|