There has been quite a bit of traffic on the RDA-L listserv which essentially became a debate on whether RDA was needed, and whether MARC was so bad after all.
I struggle with discussion about RDA and related areas of FRBR and FRAD, and have wished out loud (more than once) for a ‘Dummies guide’ that would explain it all in easily digestible chunks.
I’ve also struggled with the recent discussion on the RDA-L listserv since so much of it is along the lines of “There is nothing wrong with MARC – so what’s the point in changing”.
I have concerns about RDA, and about FRBR and FRAD – but I find the argument that we don’t need to implement them because MARC is fine just as it is depressing – if there is one thing I’m pretty sure of, it is that MARC is not OK. MARC is so not OK for where we need to be now with metadata that I don’t really know where to start. The issues with MARC have also been widely discussed on the NGC4LIB listserv, and rehashing the arguments again seemed pointless – so I didn’t bother joining in the discussion, leaving others to fight it out.
However, I did have a private correspondence with one of the participants, and I thought I ought to put some of the thoughts I expressed there into a blog post – both to air them publicly, and so I know where I’ve put them.
I should probably start by saying that I don’t regard myself as an expert in the area of metadata, so I’m quite happy to be corrected if I’ve misapprehended anything. I have to admit that I’m not even quite sure that my ‘problems’ with RDA are really actually anything to do with RDA, but perhaps more to do with how it is likely to be implemented.
I guess that the main issue that I have is that if we are really going to change the ‘silo’ nature of library data, we need a system of metadata that embraces linked data as a fundamental principal. As far as I can see, RDA does not do this. Although it does open up the possibilities of linking data, it doesn’t make it fundamental – and I believe it really needs to.
I can see that some of RDA – the work done by Diane Hillman, Karen Coyle et al on vocabularies – sets up the possibility of using linked data – but I just don’t think it is going to be enough. I’m very supportive of this work, and think it may be our best chance for RDA to realise some of its potential. However, the risk that I see is that RDA is implemented, but fundamentally not much changes.
The other problem that I have is that RDA is ‘based on’ FRBR and FRAD – and although I very much believe in the concepts behind FRBR and FRAD, I’m worried by some aspects of using them as the basis for RDA. For a start, I’m not convinced that having a conceptual model necessarily means we should bake the conceptual entities into our Resource Description rules
Secondly, I think FRBR and FRAD are OK, but I’m not sure they are really robust enough to base real world resource description on them. There are inconsistencies between FRBR and FRAD – see the discussion about ‘people’ from the DC-RDA listserv earlier this year. I think some of the things FRBR says about what counts as a separate Work are odd – e.g. two films of the same play are different works. I realise that others would disagree with me on this – which is fine, but seems an inevitable consequence of trying to apply a conceptual model in this way. Others have expressed their issues with the FRBR model in more detail and more eloquently than me – notably the work that Martha Yee has done.
I guess the way I would put this is that I believe we should create Resource Descriptions in such a way that it is possible to view them in a FRBRised way. I don’t think this is the same as starting with FRBR as a way of describing resources. I have to admit to being in two minds about this – sometimes I am convinced of the strength of the FRBR entities as fundamental to how we catalogue, and sometimes I feel that we should focus on FRBR at the presentation end, not at the resource description end. I guess that what I am sure about is that a Resource Description framework has to support the ability to display things in a FRBRized display (much more than the current situation), but I’m not sure that cataloguing in a FRBRized way is necessary to achieve this.
If these concerns all seem a bit hand-wavy and general, then I apologise – it is because they are. It could be that a good real-world implementation of RDA will overcome my concerns – but realistically I think we will see the minimum effort expended, with maximum backwards compatibility, and that will result in something that isn’t much more than MARC + AACR3 🙁
My own (vague and not at all thought through!) vision of how resource description should work is that it needs to embrace the concepts of ‘linking’ and ‘crawling’ (and is, I guess, semantic web-ish at heart). Always link when you can, and ‘crawl’ the data to build your catalogue and indexes. The way I think about it, library ‘catalogues’ would become a bit like the Google ‘copy’ of the internet – built by crawling a web of data – however, by taking advantage of the structured data available in catalogue records, it could provide more than just keyword searching.
I suspect that some of the ideas here need more thinking through, and expansion, but hopefully this is enough for now as a basis for more thought and discussion.
Owen,
I have exactly these same qualms about FRBR/FRAD and their relationship with RDA and bibliographic description. To me, the “work” is something discoverable, a relationship between resources, that could be “discovered” using different rules at different times. And the Group 1 entities in FRBR, which completely drive me nuts, should not be treated as entities but as relationships. Something is not a work, but may have a work relationship to something else. Group 1 should not define our record structure or the structure of our cataloging rules.
We need more examples, but here’s something that I can offer today: You have a book in hand. On the title page it lists someone as translator; on the verso of the title page it says: “originally published as…”. These two items have the relationship that one is a translation of the other. By someone’s definition, that makes them expressions of the same work, even if you do not know what that work actually is. (You could have a translation of a translation, for example.) For the user of bibliographic data, this relationship between the expressions is probably understandable, and possibly useful. I’m not sure that declaring something a “work,” such as in the case of two films both based on a single play, is useful; but knowing that the two films are based on the same play could be. So I think we should emphasize the relationships between resources, and let communities display these to users in the way they consider most useful to those users.
And I think we should modify FRBR to drop the Group 1 entities and redefine those as relationships. The entity should be the resource being described.
Thanks for the neat post, I agree with the spirit of what you are saying. I’m no expert on FRBR, FRAD, RDA–but there aren’t many I don’t think, and that is the main problem IMHO.
I actually think that there is a lot of potential in RDA for linked data. In particular the work that Karen is doing in the Metadata Registry http://RDVocab.info/Elements, which makes RDA into an RDF vocabulary http://RDVocab.info/Elements.rdf is really exciting.
Some people might moan-and-groan about RDF, but it’s pretty much the state of the art for metadata deployment on the web. Alistair Miles’ involvement in RDA is also really key, since he definitely understands what linked data and the web are all about. http://alimanfoo.wordpress.com/2007/05/02/rda-resource-description-and-access/
This post originally came out of a correspondence with Diane Hillman, and she emailed me her response to much of what I posted here. With her permission I’ve posted her response here, with extracts from my original email:
> OK, what are my problems with RDA? I should probably say that I don’t
> regard myself as an expert in the area of metadata, so I’m quite happy
> to be corrected if I’ve misapprehended anything 🙂 I have to admit that
> I’m not even quite sure that my ‘problems’ with RDA are really actually
> anything to do with RDA, but perhaps more to do with how it is likely to
> be implemented.
>
Implementation, of course, is likely to be messy and is certainly not
able to be “managed” or “policed” by anyone. I suppose, even though
market-driven decisions are currently not in good odor, that’s what
we’re stuck with. That said, we can certainly do things to encourage
appropriate implementation, though I’m not always sure what those things
might be, and whether the small group working in this is able to do
those things (and if not us, who?). Suggestions always welcome …
> I guess that the main issue that I have is that if we are really going
> to change the ‘silo’ nature of library data, we need a system of
> metadata that embraces linked data as a fundamental principal. As far as
> I can see, RDA does not do this. Although it does open up the
> possibilities of linking data, it doesn’t make it fundamental – and I
> believe it really needs to.
>
>
Yup, I agree, and there was a time that Karen and I seriously considered
becoming the RDA competition rather than the people trying to herd, push
and enable RDA to be “the best it can be,” or at least, as you note, be
accommodating of the kind of data we think is necessary for real
progress. As it is, we’re part way there, and it remains to the
implementers to push that envelope a bit, as we move forward.
> I can see that some of RDA – and the work that you and Karen Coyle have
> done with vocabularies – sets up the possibility of using linked data –
> but I just don’t think it is going to be enough. The risk that I see is
> that RDA is implemented, but fundamentally not much changes.
>
>
Well, again, we made a choice to go with what the JSC came up with for
the first pass, on the theory that the best way to get implementers
trying it and building up sufficient experience to recommend fixes for
the stuff that didn’t work was to get it out the door and into the hands
of people like you. That’s the iterative model, no? We figured that
the alternative was another five years or so of discussion, or somebody
having to start from scratch because there was nobody willing to try out
the admittedly non-perfect first pass. I, for one, don’t have the
patience for that, nor maybe even the time on earth (I’m one of Bill and
Hillary’s generation, after all). It will, I hope, be enough to start,
then we’ll have to figure out how to improve it until it’s “good
enough.” You have a better idea?
> The other problem that I have is that RDA is ‘based on’ FRBR and FRAD –
> and although I very much believe in the concepts behind FRBR and FRAD,
> I’m worried by some aspects of using them as the basis for RDA. For a
> start, I’m not convinced that having a conceptual model necessarily
> means we should bake the conceptual entities into our Resource
> Description rules
>
>
Well, again, we don’t have a crystal ball, either, but FRBR/FRAD are a
hell of a lot better than the “we don’t need no model” development we’ve
been used to, don’t you think? I think we’re well down that road, and
we’ll have to try it out, see if it floats. I doubt whether another
decade or so of discussion about whether they should or shouldn’t be the
basis for how we think about description will get us much further.
> Secondly, I think FRBR and FRAD are OK, but I’m not sure they are really
> robust enough to base real world reseource description on them. There
> are inconsistencies between FRBR and FRAD – see the discussion about
> ‘people’ from the DC-RDA listserv earlier this year. I think some of the
> things FRBR says about what counts as a separate Work are odd – e.g. two
> films of the same play are different works. I realise that others would
> disagree with me on this – which if fine, but seems an inevitable
> consequence of trying to apply a conceptual model in this way. I guess
> the way I would put this is that I believe we should create Resource
> Descriptions in such a way that it is possible to view them in a
> FRBRised way. I don’t think this is the same as starting with FRBR as a
> way of describing resources. I have to admit to being in two minds about
> this – sometimes I am convinced of the strength of the FRBR entities as
> fundamental to how we catalogue, and sometimes I feel that we should
> focus on FRBR at the presentation end, not at the resource description
> end. I guess that what I am sure about is that a Resource Description
> framework has to support the ability to display things in a FRBRized
> display (much more than the current situation), but I’m not sure that
> cataloguing in a FRBRized way is necessary to achieve this.
>
>
Oh yeah, there are lots of inconsistencies, but whether we had the time
to iron those out prior to rolling out RDA is quite another thing.
Gordon Dunsire from Strathclyde is working with IFLA to register the
FRBR entities and relationships, and he’s picking out a lot of those
inconsistencies. I contend that without the concrete necessity to
create these formal representations, we’d never surface most of those.
I agree with you that we could create FRBRized views of just about
anything–OCLC has done that fairly nicely with Fiction Finder. But I
contend that unless and until we start thinking in FRBR
terms–particularly about relationships IMHO (which we won’t do until we
have to do it as part of creating descriptions), we’ll never get it
right, and never figure out how to make it work. I would also suggest
that we will NEVER, EVER agree on what’s a work and what’s not in every
situation for every community of practice, and what we need to do
instead is find a way of expressing how a community describes their
works, expressions, etc., and be able to display other ways to look at
the same description. So, as an example, if you look at the DCMI/RDA WG
wiki, and the wonderful scenario (#4, I think) that Greta deGroat
contributed, you’ll see immediately that how she looks at her materials,
and how she proposes to describe them in FRBR terms, will not work very
well for public libraries, who need a much more generalized view. We’ll
need to figure out how to do that–I personally think that Application
Profiles is the way to go.
> If these concerns all seem a bit hand-wavy and general, then I apologise
> – it is because they are. It could be that a good real-world
> implementation of RDA will overcome my concerns – but realistically I
> think we will see the minimum effort expended, with maximum backwards
> compatibility, and that will result in something that isn’t much more
> than MARC + AACR3 🙁
>
>
Well, I think we’ll have to see. I know that even in the last 6 months
or so there’s been lots more interest in this stuff on the part of
implementors, so I tend to remain optimistic. In the short term, it’ll
be messy and bloody, but when have we seen anything else when big
changes are afoot?
> My own (vague and not at all thought through!) vision of how resource
> description should work is that it needs to embrace the concepts of
> ‘linking’ and ‘crawling’ (and is, I guess, semantic web at heart).
> Always link when you can, and use crawling to build your catalogue and
> indexes. The way I think about it, library ‘catalogues’ would become a
> bit like the Google ‘copy’ of the internet – built by crawling a web of
> data.
>
>
Well, yes, that’s very semantic-webby of you, and far be it from me to
discourage that, but libraries need some kind of transition to bridge
the rather enormous gap between what they do now and what you’re talking
about. The good news, I think, is that all the formal representations
are happening using SemWeb friendly technology, so when the time comes
for you to run the world, we’ll be able to toss you a few billion
triples to play with, and they should be pretty meaty. For now, we’ve
got a lot of work to do to get to that point.
So, let me know when you’re ready to talk turkey about implementing RDA
… 🙂
Just to comment in turn on the comment from Diane above.
I wouldn’t want to be a brake on the process of getting RDA out of the door, and I agree generally that getting RDA delivered now is better than discussing it for another 5 years. Release early, release often as they say 🙂
I also would stress the admiration I have for the work that Diane and others have done in opening up the possibility of using linked data in the context of RDA – this is invaluable.
As I struggle to see what the key steps are, I become more convinced that opening up the ability to link data is the key step for library metadata, and this is the most important thing we can achieve here. As such, RDA is not a bad start, and although I may have concerns over some of the detail, perhaps these need to be put aside for the time being to allow us to move forward.