-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

[Apologies again for cross-posting, but it was just suggested to me
that XML4LIB and PERL4LIB probably won't cover the possible audience]

The Perl module Net::OAI::Harvester implements a client framework for
the OAI Protocol for Metadata Harvesting (OAI-PMH) and was authored
and originally maintained by Ed Summers. It has been available on
CPAN ever since 2003 and its last stable version 1.15 has been
released almost four years ago:
         < http://search.cpan.org/~thb/OAI-Harvester-1.15/ >.

Since one of the repositories used for testing vanished from the web
some time ago and this is breaking the test suite a new version has
to be released fairly soon.

Over time I had been tackling various minor issues and published
developer releases on CPAN, cf. the list at the end of this mail
or the Changes document linked at the CPAN page for the current
release 1.16_12: < http://search.cpan.org/~thb/OAI-Harvester-1.16_12/ >

However the sum of these changes is not negligible and specifically
their impact on "custom metadata handlers" (which are to be used when
processing other metadata formats than oai_dc) may affect
applications using the module:

>>>>>
Up to version 1.15 the metadataHandler was inconsistently fed with input
:

 - GetRecord exposed the almost complete XML response to the Handler
   (including start_document/end_document events)
 - ListRecords exposed the (OAI)record element (header, metadata and
   optional about containers) but did not propagate start_document or
   end_document events.

In both cases the events for the header tags itself and for the optional
setSpec subelements had not been forwarded

Version 1.20 introduces a modified behavior for metadataHandler and an
additional recordHandler:
 - a metadataHandler will see only the (single) subelement of the OAI
   metadata element (so for an deleted record it might never be invoced
   at all)
 - a recordHandler will see the OAI record element and its subelements

Therefore a metadataHandler will now be confined to the metadata
fragment(s) of the response, and the new recordHandler approximates the
old behavior of ListRecords, however OAI-PMH:identifier and
OAI-PMH:datestamp will now be properly encapsulated within their
OAI-PMH:header element.

Additionally, two new methods responseDate() and request() allow access
to the corresponding top-level OAI-PMH elements in all response types.
A SAX filter of class Net::OAI::Record::DocumentHelper may be used to
inject start_document and end_document events into the chain if they
are needed.

As a temporary measure, you may set
   $Net::OAI::Harvester::OLDmetadataHandler =1
to change the behavior of handlers passed as "metadataHandler" into
that of a recordHandler.
<<<<<<

Obviously the change of semantics for a metadataHandler to deal with the
"metadata" elements of the response instead of the "record" elements
is a design decision and may be questioned by users of the module.

The current version also contains several changes which solve
deficiencies of Net::OAI::Harvester 1.15 but possibly break existing
workarounds for these deficiencies. For example officially (per
documentation) you never could acccess
the responseDate of the OAI-PMH result, but due to a sloppy
implementation of processing for the identify verb it was possible to
extract it in this case by an undocumented method. The current version
supplies a dedicated responseDate() accessor for all verbs but at the
same time fixes the behavior in the identify case.

I may be overly optimistic but my impression is that the changes between
the current 1.15 and the coming version (most probably numbered 1.20)
do actually fix many issues but the fear is realistic (I experienced
that myself with an old application of mine using the module) that
these fixes may conflict with workarounds introduced by users to make
things work before.

 ***

So please, if you are currently using Net::OAI::Harvester *and* had
been forced to introduce workarounds or tweak internals of the module,
perform thorough testing before upgrading to the coming stable
version, preferably already now with the developer version 1.16_12.

And, please, please: provide feedback if you should run into trouble,
either via the CPAN request tracker for the module at
< https://rt.cpan.org/Public/Dist/Display.html?Name=OAI-Harvester >
or by direct mail.

Sorry for the inconvenience
viele Gruesse
Thomas Berger


Changes to Net::OAI::Harvester since version 1.15

1.16_12	Tue, Jan 12 00:20:05 CET 2016
- - dealing with CPANTS Kwalitee issues, esp. version number mess
- - new filter class Net::OAI::Record::DocumentHelper for tweaking

1.16_11	Tue, Jan 12 00:20:05 CET 2016
- - minor cleanup

1.16_10	Mon, Jan 11 01:29:46 CET 2016
- - renamed alldata() method for accessing recordHandler results
  to recorddata()
- - better propagation of namespace prefix mapping events
- - Net::OAI::NamespaceFilter with a result() method
- - Net::OAI::NamespaceFilter tested with XML::SAX::Writer
- - AUTHOR formatting

1.16_09 Sun, Feb 14 17:29:39 CET 2014
- - Net::OAI::NamespaceFilter as kind of generic metadata handler
- - Queries are now constructed basing on a copy of the Harvester's
  baseURL
- - pass parameters to URI->query_form() more reproducably,
  (esp. "verb" should now always be first to accommodate some
  allegedly broken repositories)
- - temporary? tests for correctness of LWP operations

1.16_07	Tue, Apr 30 01:26:40 CEST 2013
- - added new methods: response(), responseDate(), error()
- - Smoke still tests failed on 'Bad Host' tests (wrong error codes
  induced by HTTP proxies?)
- - aligned behavior of metadataHandler for listRecords() and
  getRecord()
- - introducing alternative recordHandler for listRecords() and
  getRecord()
- - removed erroneous resumptionToken handling for identify()

1.16_04	Fri Dec  7 09:49:03 CET 2012
- - consider HTTP proxies in design of t/003.error.t
- - 'Bad Host' tests failing b/c error code 500 is not the expected
   code 404 (due to some recent change in LWP)?

1.16_01 Mon Apr  2 23:14:35 CEST 2012
- - Modules were not namespace aware.
- - Add HTTPRetryAfter() method (catches HTTP Retry-After header)
- - Check responses for Content-Type and charset before parsing
- - Net::OAI::Header handed up (empty) header elements and other stuff
  to the request's metadataHandler
- - SKIP tests when HTTP errors are encountered

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1
Comment: Using GnuPG with Thunderbird - http://www.enigmail.net/

iJwEAQECAAYFAlaZJ9QACgkQYhMlmJ6W47MK8wP+P0sxqbIO5F9mofNkyDUHT7Ky
rWmOkg4lK5Rc63AOMporyZps7jbPrHPaMReWAxfT8+PF+XbrzWMBqIb9j8SPrPpK
FIZ8ghXtv/P3btlGsv/aI+Tt9EtGTAvDJYU+XLiuEy5qn3VNC8jyTSywbeSWV7B4
GhOOhnKpQHb42THYRaY=
=ajt0
-----END PGP SIGNATURE-----