This generally agrees with what I have found[1]. That is, if a second
indicator of '0' exists, then it is fairly likely the URL will lead you to
the full item. However, unfortunately an indicator of '1' is ambiguous, so
it's possible you will be leaving a lot behind if you rely on that solely.
It sure makes for easier code, though, no doubt.

Also, if people were more interested generally in what is in various 856
subfields for book/journal MARC records, they can refer to my "MARC Usage
in WorldCat" work[2]. It can be amusing, at least, if not instructive.


On Mon, Feb 17, 2014 at 9:45 AM, Jonathan Rochkind <[log in to unmask]> wrote:

> Interesting, thanks for the additional information, very useful!
> I don't like relying on the 'free text' in subfield 3, because it seems
> fragile, who knows if I know all the possible values or if they change them
> in the future breaking my code.
> But your example with two 'full text' links is enlightening.
> I think what I'm liking as an algorithm for my needs (any full text is
> better than none, but PDF is best) -- is first looking for an 856 with
> second indicator "0" -- if there's only one, use it. If there are more than
> one, try to find one that includes the substring "PDF", if none do, just
> use the first one.
> Jonathan
> On 2/17/14 11:16 AM, Andrew Anderson wrote:
>> The document you want to request from ProQuest support was called
>> Federated-Search.docx when they sent it to me.  This will address many of
>> your documentation needs.
>> ProQuest used to have an excel spreadsheet with all of the product codes
>> for the databases available for download from
>> source=article&c=12&cid=26, but it appears to no longer be available
>> from that source.  ProQuest support should be able to answer where it went
>> when you request the federated search document.
>> You may receive multiple 856 fields for Citation/Abstract, Full Text, and
>> Scanned PDF:
>> =856  41$3Citation/Abstract$u
>> =856  40$3Full Text$u
>> =856  40$3Scanned PDF$u
>> I would suggest that rather than relying on the 2nd indicator, you should
>> parse subfield 3 instead to find the format that you prefer.  You see the
>> multiple 856 fields in the MARC records for ProQuest holdings as well, as
>> that is how ProQuest handles coverage gaps in titles, so if you have ever
>> processed ProQuest MARC records before, you should be already prepared for
>> this.