z3950 search and import seems to works fine.
Let me explain how :
* a "search z3950" button is added in the addbiblio template.
* when clicked, a popup appears and z3950/search.pl is called
* z3950/search.pl calls addz3950search in the DB
* the z3950 daemon retrieve the records and stores them in z3950results AND in marc_breeding table.
* as long as there as searches pending, the popup auto refresh every 2 seconds, and says how many searches are pending.
* when the user clicks on a z3950 result => the parent popup is called with the requested biblio, and auto-filled
Note :
* character encoding support : (It's a nightmare...) In the z3950servers table, a "encoding" column has been added. You can put "UNIMARC" or "USMARC" in this column. Depending on this, the char_decode in C4::Biblio.pm replaces marc-char-encode by an iso 8859-1 encoding. Note that in the breeding import this value has been added too, for a better support.
* the marc_breeding and z3950* tables have been modified : they have an encoding column and the random z3950 number is stored too for convenience => it's the key I use to list only requested biblios in the popup.
* adding a syntax column in z3950 table = this column will say wether the z3950 must be called with PerferedRecordsyntax => USMARC or PerferedRecordsyntax => UNIMARC. I tried some french UNIMARC z3950 servers, and some only send USMARC, some only UNIMARC, some can answer with both.
Note this is a 1st draft. More to follow (today ? I hope).
It was due to an illegal contruction in Koha : we tried to retrive subfields from <10 tags.
That's not possible. MARC::Record accepted this in 0.93 version, but it was fixed after.
Now, the construct/retrieving is OK !
