Despite some issues, ISBNs very often reliably identify two distinct Instances of a Work, both as defined in the BIBFRAME model.  The resulting descriptions of Instances can be quite diverse:

# The Work
<> bf:creator <>;
	bf:label "Card, Orson Scott. Ender's game";
	bf:language <>;
	bf:title "Ender's game";
	a bf:Book,

# Instance 1
<> bf:dimensions "22 cm.";
	bf:extent "357 p. ;";
	bf:instanceOf <>;
	bf:isbn10 "0312932081";
	bf:isbn13 "9780312932084";
	bf:placeOfPublication "New York, NY";
	bf:pubDate "1985";
	bf:publisher "T. Doherty Associates";
	a bf:HardCoverBook,

# Instance 2
<> bf:dimensions "6.7 x 4.2 x 1.1 inches";
	bf:extent "352  p. ;";
	bf:instanceOf <>;
	bf:isbn10 "0812524586";
	bf:isbn13 "9780812524581";
	bf:pubDate "1986";
	bf:publisher "Arrow Books";
	a bf:Instance,

That said, we also recognize that the publisher and physical description information in a MARC record with multiple ISBNs pertains to only one of those ISBNs.  In the case of a MARC record with two ISBNs, one for the hard back and one for a paperback, a decision will have to be made which ISBN the publication and physical description information should be associated with.  For us, this is pretty easy - the publisher and physical description almost always refers to the hardback.  This will result in some skimpy Instance resources, but that does not invalidate the principle.  Further, it actually creates divisions in the data where one would want divisions.  Depending on the scenario, an enterprising cataloger - holding the paperback version - may discover a pre-existing Instance resource based on the ISBN (and its relationship to the correct Work) and record a fuller description of the Instance (for the paperback).

We are very much aware of the cataloging practices in the past.  We're looking forward to a bibliographic ecosystem of the future, where we hope to see a greater demarcation between things.


> -----Original Message-----
> From: Bibliographic Framework Transition Initiative Forum
> [mailto:[log in to unmask]] On Behalf Of Karen Coyle
> Sent: Friday, May 17, 2013 11:23 AM
> To: [log in to unmask]
> Subject: Re: [BIBFRAME] What's an instance?
> Kevin, what people are telling you is that there is a good reason why
> libraries do not create separate records for things like hardcopy v.
> trade paperback, and they would not want separate instances for these.
> This is a decision that was made many, many decades ago. The data that
> you will have in those MARC records will be identical for the two ISBNs,
> which means that the record does not describe both. You only have one
> instance description.
> kc
> On 5/17/13 7:27 AM, Ford, Kevin wrote:
> > Thanks, Jörg.
> >
> > One very quick point:
> >
> >> Do not rely on ISBN as a unique
> >> identifier.
> > -- Just to be absolutely clear, we're not trying to use an ISBN as an
> identifier.  Indeed, relating Instances that use the same ISBN is not
> even in our view at present (and may never be, for all the reasons
> given about ISBNs to date).  Likewise, there are no plans to relate
> Works that have Instances that share ISBNs.  But, if we have a MARC
> record with three different ISBNs in it, as a general rule, we are
> making the assumption that each represents a distinct Instance of the
> Work being described (in our records there are often two ISBNs - one
> for the hard cover and the other for the paper back).  Now, the
> calculus will change if we can reasonably determine we are dealing with
> a MARC record that describes a multi-volume Work, for example.  The
> three ISBNs could mean something different in such a scenario.
> >
> > *Per MARC record*, ISBNs offer a guide to splitting said MARC record
> into a BIBFRAME Work and one or more BIBFRAME Instances.
> >
> > Yours,
> > Kevin
> >
> >
> >
> >
> >
> >> -----Original Message-----
> >> From: Bibliographic Framework Transition Initiative Forum
> >> [mailto:[log in to unmask]] On Behalf Of Jörg Prante
> >> Sent: Friday, May 17, 2013 9:19 AM
> >> To: [log in to unmask]
> >> Subject: Re: [BIBFRAME] What's an instance?
> >>
> >> Kevin,
> >>
> >> yes, I can run an analysis over a snapshot of our union catalog
> >> comprising of around 4.847.212 ISBNs. It will take some time to
> write
> >> the code.
> >>
> >> Due to the official policy of the ISBN agencies, the reuse was
> >> strictly forbidden - but the ISBN number pool is segmented into
> >> countries and their publishers, which take charge over the number
> >> use. There is no technical mechanism to enforce correct use or to
> >> grant or revoke ISBNs by a third party. Beside applying the same
> >> to different editions, there are other cases. Some publishers wanted
> >> to save resources and simply broke the ISBN rules when they ran out
> >> of money (or they were rejected from receiving more ISBN numbers).
> >> They started to recycle ISBNs of books they had out of print for
> many
> >> years, hoping no one will ever notice. Or, publishers did not
> properly file their ISBN pool usage.
> >> For example, when publishers took over other publisher's business
> and
> >> their ISBN pools, there was no safe way of verifying what ISBNs were
> >> already taken or not. ISBN was in active use as primary identifiers
> >> in ISBN registers for no longer than 5 or 10 years. It's giving us
> >> headaches for a very long time. Do not rely on ISBN as a unique
> >> identifier.
> >>
> >> Jörg
> >>
> >> Am 16.05.13 23:54, schrieb Ford, Kevin:
> >>>    Is there any way to quantify, for example, how often publishers
> >> actually reuse ISBNs in different editions (is that even tecnically
> >> permitted?)?
> --
> Karen Coyle
> [log in to unmask]
> ph: 1-510-540-7596
> m: 1-510-435-8234
> skype: kcoylenet