I think one thing that's often overlooked about the Umlaut and was a
constant source of frustration for me, in the early days, which is now
Jonathan's burden, is that the Umlaut was designed specifically for
two purposes:

1) Reduce the extremely high percentage of failure in identifying
known and locally (or consortially) available resources in the link
resolver chain
  a) The main use case here was conference proceedings, but wound up
being broadly applicable to all sorts of things in the library's
collection, you know, like books.
2) Provide access to items not physically (or electronically) in the
library's collection but have a freely available surrogate on the web
(pre/post prints/digitized copies/illegal copy on person's web site
crawled by Google/etc.).

#1 has sort of a solution via services such as Summon and its ilk - if
you have a fulltext index of everything your library owns or
subscribes to, it makes it a lot easier to find things.  Having them
all in the same bucket simplifies this even more.

This, in my mind, isn't a great approach, but it at least can address
that particular problem.

#2, however, has been largely ignored.  While you might get
centralized repositories like HathiTrust, or Citeseer in a
system like Summon or EDS (although I don't actually see the latter
two listed in their sources lists), Google Scholar shows just how much
literature is available outside of any controlled silo.

The fact that libraries have, so far, completely ignored this freely
available content (and let's face it, the fact that not all of it
*should* be freely available is neither here nor there - if it's the
article you need and it's found via simple Google-ing, it's game) and
have constrained the discovery/delivery process to the tiny percentage
of the universe that they own or lease is mind blowing and completely
antithetical to what libraries are supposedly about.

Karen, you're right - the Umlaut is not going to be viable most
libraries (case in point: Georgia Tech, an ARL library, didn't feel
that they could maintain it after I left and shut it down), but that
was never its intention.  The point of the Umlaut was not for hundreds
or thousands of libraries to run Umlauts, but to show that a single
person, toiling in the basement of a single library (originally in
Georgia, now in Maryland) with a single application can solve all of
these problems that our vendors are apparently unwilling or unable to
do.  The simplest parts of the Umlaut could be automated, the harder
bits could be abstracted a bit and crowdsourced.  I'm not aware of any
part of Umlaut (besides it's x-identifier stuff) that has trickled
back into a vendor product.  I don't know of any library project,
commercial or otherwise, that is using Google, Yahoo! or Bing (and,
granted, neither is Umlaut anymore).

The stuff the Umlaut does isn't hard, but it requires somebody to
*care* and be dedicated enough when things change or break to tweak
things as necessary.

The fact that libraries find this to be *too much work* is just
absolutely demoralizing to me.


On Wed, Jun 15, 2011 at 12:07 PM, Jonathan Rochkind <[log in to unmask]> wrote:
> On 6/15/2011 9:31 AM, Eric Hellman wrote:
>> Clearly, Jonathan has gone through the process of getting his library to
>> think through the integration, and it seems to work.
> Thank you!
>> Has there been any opposition?
> Not opposition exactly, but it doesn't work perfectly, and people are
> unhappy when it doesn't work. It can sometimes find the _wrong_ match on a
> 'foreign' site like Amazon etc.  Or avoid finding a right one of course.
> Or the definition of right/wrong can be not entirely clear too -- on a bib
> record for a video of an opera performed, is it right or wrong to supply a
> link to the print version of the opera? What if the software isn't smart
> enough to _tell_ you it's an alternate format (it's not), and the link is
> just in the single flat list of links?
> Also issues with avoiding duplicate double URLs when things are in bib
> records AND in SFX kb AND maybe looked for otherwise by Umlaut. (we have
> _some_ HathiTrust URLs in our bib records, that came that way from OCLC, who
> knew?)
> These things get really complicated, quickly.  I am constantly finding time
> to do more tweaking, but it'll never be perfect, so people have to get used
> to lack of perfection. Still when I ask, okay, this HathiTrust/Amazon/Google
> linking feature is not going to be perfect, would you rather keep it with
> imperfections we may not be able to fix, or eliminate it -- nobody says
> eliminate.
>> What are the reasons that this sort of integration not more widespread?
>> Are they technical or institutional? What can be done by producers of open
>> access content to make this work better and easier? Are "unified" approaches
>> being touted by vendors delivering something really different?
> I think they are mostly technical.  This stuff is _hard_, because of the
> (lack of) quality of our own metadata, the lack of quality of third party
> metadata, the lack of sufficient APIs and Services, and the lack of a local
> technical infrastructure to support tying everythign together.
> So on the one hand, I'm trying to find time for an overhaul of Umlaut to
> make it easier for people to install and maintain, and I'm hoping I can get
> some more adoption at that point.  To at least provide some open source
> "local technical infrastructure". Umlaut is intentionally designed to be as
> easy as possible to integrate with your existing catalog or other service
> points, as well as to provide 'just in time' services from third party
> external searches -- that's it's mission, this kind of just-in-time service.
> ("easy as possible" -- or as easy as I can make it, which sometimes still
> isn't easy enough, especially if you don't have local technical resources).
> But still, it's metadata, metadata, metadata.  So what can producers of open
> access content do to make this work better and easier?
> 1) Have good metadata for their content, especially including as many
> identifiers as possible -- ISBN, OCLCnum, LCCN.   Even if you aren't an OCLC
> member and don't have an "OCLC record", if you can figure out what OCLC
> record represents this thing you've got, list it in the metadata.  Even if
> the ISBN/OCLCnum/LCCN doesn't represent the _exact_ same thing, list it --
> ideally somehow identified as 'an alternate manifestation'.  Also have
> author, title, publisher, publication year  metadata.  If you can have
> author metadata as an NAF/VIAF controlled form or identifier, even better.
>  Metadata is expensive, but metadata is valuable, the better it is, the
> better Umlaut's approach can work.
> Share the metadata publically, in case someone wants to do something with
> it.
> 2) Provide an API that allows lookup of your open access content, searching
> against the good metadata from #1. Including identifier searches.  The thing
> is, each of (dozens, hundreds, thousands) of open access content providers
> having such an API --- it's a burdensome expense for each of them, but it's
> also unrealistic for client software to talk to dozens/hundreds/thousands of
> APIs.
> So this stuff needs to be aggregated in fewer major service points.  It
> could be an aggregator of just metadata that links to content hosted on
> individual hosts, or it could be an aggregator of content itself. Either
> way, it needs a good API based on good metadata. "Google" doesn't work as
> such an aggregator, the APIs it has are too limited functionally and by ToS,
> and the results do not have sufficient metadata.  Maybe the Internet Archive
> does -- although IA's API's and metadata are sometimes a bit sketchy (If you
> do put it in IA, make sure it somehow shows in the "Open Library" section
> and it's APIs -- the OL API's IA has are sufficient for Umlaut's use, but
> general Internet Archive APIs are not).  Or maybe a new aggregator(s) have
> to be collectively created.