[Koha-devel] Zebra and Directory Structures

[Paul POULAIN]

Joshua Ferraro a écrit :
> Hey guys,
> 
> I'd like to propose that we reflect for a bit about how to structure the
> zebra components within rel_3_0 before we diverge too much. 

good proposal.

> There are currently three zebra databases for a given rel_3_0 install, 
> and it's likely there will be more in the future (if we have a metarecord
> bibliographic for instance), but for now we have:
> 
>   biblios (biblio records)
>   authorities (authorities records )
>   public (for a publicly accessible Z39.50)
> 
> Each of these needs its own 'tab' directory, as well as registry
> locations, etc (for the database files).

??? I thought public was just for a public access to biblios database & 
was not requiring more than a definition in the config file. Could you 
detail a little bit more what it is done for ?

> We also have at least two MARC types to deal with, with more to follow
> if our plans for world domination succeed :-): usmarc and unimarc.
> Each flavor of MARC needs it's own tab files.

not only tab files, but also record.abs, ccl.properties, ...

> I propose we create a standard for managing the zebra files as follows.

very good suggestion !

> The installer creates one directory 'zebra' with the following contents:
> 
> $ ls zebra/
> unimarc usmarc
> 
> then, in each of these we'll have:
> 
> $ ls zebra/usmarc/
> authorities biblios public
> 
> within each of these we'll have in turn:
> $ ls zebra/usmarc/biblios/
> key  lock  recordDelete  register  shadow  specialUpdate  tab  tmp

why do we need to have this in CVS ? I suggest we could have a script 
that build all of these for you when setting up koha.

and, you know what ? I already did a large part of that script : 
misc/migration_tools/rebuild_zebra : it create all necessary 
directories, copies all needed files, all of this depending on your 
koha.xml file.
then it exports & reindex everything !

It has to be completed & tweaked but afaik it works correctly.

> This way, if I have one server with three disks, I can have the
> following distribution:
> 
> Filesystem	Mounted on
> /dev/sda1 	/
> /dev/sda2 	/koha/zebra/usmarc/biblios
> /dev/sda3 	/koha/zebra/usmarc/authorities
> 
> (ie, in this way, zebra biblios, authorities and mysql don't compete
> for disk usage).

I'm almost OK with this structure, even if I think we don't need the 
usmarc level, just a library level, to be able to have X koha on the 
same server :
/koha/zebra/library1/biblios
/koha/zebra/library2/bilbios
/koha/zebra/library1/authorities
/koha/zebra/library2/authorities

I think it's more important to have everything for a given library in 
the same directory structure than having a disk dependant directory 
structure.

> I also propose we attempt to keep unimarc and usmarc in as close sync as
> possible, using the same naming conventions, etc., to make sure we
> don't get confused. (I noticed that in the unimarc directory, there
> are directories with capital letters as the first letter of the dir ...
> which is different than in usmarc ... we should pick one method and
> stick with it )

kados +++++++++++

I already have problems with Koha-number, Local-Number, 
Identifier-standard ...

> All of the non-tab files (like koha.xml, ccl.properties, pqf.properties,
> zebra-biblios.cfg, zebra-authorities.cfg, zebra-public.cfg should IMO
> go into the koha etc directory, 
mmm... for multiple setups, do will be use /etc/koha/setup1/... 
/etc/koha/setup2 /etc/koha/setup3 ... ?
or will we prefer /home/zebra/library1/etc to have a single place for 
every library1 related things.
I'me not found of one solution, so you'r will be mine, I just wanted to 
point the question.
> but in CVS we should have:
> misc/etc/usmarc/
> misc/etc/unimarc/
> because the files are different for each flavor of MARC. Or should we
> store these config files in the tab dir?

mmm... I'm not sure, but I think we should have a different dir for all 
of them, not a record_for_unimarc.cfg as we have actually.

> Finally, once we decide on a method, lets clean up CVS by deleting all
> the extra directories and files (like zebraplugin and misc/zebra).
yep.

-- 
Paul POULAIN et Henri Damien LAURENT
Consultants indépendants
en logiciels libres et bibliothéconomie (http://www.koha-fr.org)
Tel : 04 91 31 45 19