Re: [Firefly-dev] About the xml way

[firefly-dev]

Hello, In CVS I am not sure how much room we have. I know we should have
alot though so feel free to add the directory structure to the test
folder. I know some of what I am about to say has been covered but I
need to let everyone else hear what has been communicated to Gimpy.
This project comes from a frustration of working with bad library
software everyday. I get tired of writing cryptic MARC files.(MARC is
what most libraries use). I want a DB that is independent of its
platform, I also want software that a moron can work. I know that not
every library has a Sysadmin but most have a librarian who tried to
figure it out. I would rather come up with something that has a small
learning curve as apposed to making someone learn any of the sql's.
There is already a free library system that runs off sql, Koha. I do not
like the way Koha works and I would like to avoid being stuck in one
possition for the rest of this things life. If we figure out our own DB
plan and how to store our own files in a format that is easy to read and
write to than this will be easily adapted by libraries. Librarians most
the time don't have alot of time to sit and learn a bunch of new stuff
just so they can get their new system working, if they have something
with a small learning curve that they can install on just about any
platform be it, Sparc to ppc to x386. I want to see this be the best
thing out there for libraries. If giving up alittle speed for alittle
easier learning curve it is worth it.
John Hornbeck 


On Wed, 2003-04-02 at 04:01, address@hidden wrote:
> Hi,
> 
> I thought about the idea of gimpy in his last msg and I have to say that
> it is really appealing.
> My only concern is about the efficiency of this system, however since we
> have never tested it we can't know how much it can be slow, so I thought
> we should write some benchmark tests.
> We just need to write xml files like the patron.xml in the cvs, put them
> in a server and then read/parse/write them using small scripts in perl
> *and* in python. 
> Execute tests from another computer would be better to test the net
> overhead, but for now maybe we can avoid.
> 
> I'm thinking about the directory structure proposed by gimpy, 
> we could have something like this (just an idea for now):
> 
> /libraries
>       /library01
>               /patrons
>                       /patron_id  (lastname isn't unique)
>                               loans.xml
>                               patron.xml
>                               ....
>               /staff
>                       /staff_id
>               /catalog
>                       /books
>                               book_id.xml (a big library here has             
>                              thousands files)
>                       /newspapers
>                       /cdrom
>                       etc...  
>       /library02
>               etc...
> 
> This is very basic but I think it is enough for tests, however we should
> put lot of files to have a fair benchmark.
> 
> The tests I've done with db have been done on 100.000 records, and it
> tooks about 0.3 sec to retrieve all books info, this without network
> overhead since it was all in my pc.
> 
> If you agree this evening I'll add the directory structure in the tests
> folder with some xml file randomly generated.
> 
> BTW how much space do we have on the cvs ?
> 
> Marco
> 
> 
> 
> _______________________________________________
> Firefly-dev mailing list
> address@hidden
> http://mail.nongnu.org/mailman/listinfo/firefly-dev
-- 
John Hornbeck <address@hidden>