program using the GenTech model
Jesper Zedlitz
ago but the thread is too old to answer.
I always played with the thought to write such a program. On other technical
genealogical mailinglists there are several other persons who have similar
plans.
Ok, it is a challenge -- probably nothing for one or two persons. But maybe
we can find a group of motivated people and start developing a program that
can handle more than "person, family, note".
Right know I do not have a precise plan how to start the project. One
possibility is a top->bottom approach taking a look at several real-world
documents and see how they can be managed with the new program. Or we could
start bottom->top and design classes based on the data model and start
working with those.
My preferred programming language is Java. Geneapro used Eclipse's RCP to
build an user interface. But I am not sure if maybe a webbased solution
would be better. It would be the best to structure the project so well that
it is easy to have several different "view" layers.
To get acceptance from the users the program must hide the complexity of
Gentech in most cases. It should only be visable in 'expert mode'. So we
need a mapping between the simple 'traditional' model ("My grandmother's
name was ... and she was born on ...") and Gentech's evidence/assertion
model.
Jesper
--
Jesper Zedlitz E-Mail : jes...@zedlitz.de
Homepage : http://www.zedlitz.de
ICQ# : 23890711
Mat
> There has been a thread about the GenTech data model on this list some time
> ago but the thread is too old to answer.
I thought about replying to this about a month ago, but haven't got
round to putting my thoughts down in writing until now.
> I always played with the thought to write such a program. On other technical
> genealogical mailinglists there are several other persons who have similar
> plans.
>
> Ok, it is a challenge -- probably nothing for one or two persons. But maybe
> we can find a group of motivated people and start developing a program that
> can handle more than "person, family, note".
>
> Right know I do not have a precise plan how to start the project. One
> possibility is a top->bottom approach taking a look at several real-world
> documents and see how they can be managed with the new program. Or we could
> start bottom->top and design classes based on the data model and start
> working with those.
>
> My preferred programming language is Java. Geneapro used Eclipse's RCP to
> build an user interface. But I am not sure if maybe a webbased solution
> would be better. It would be the best to structure the project so well that
> it is easy to have several different "view" layers.
I've had a fiddle with creating something based on the GenTech GDM
model in C#, but the language isn't important. I got a certain way
forward, but there's lots of issues to deal with.
I've considered a web interface as it provides a very low barrier to
entry for casual users. But even with AJAX and the like, I don't think
a web-based interface is suitable for serious use. I suppose a
read-only web view would be nice though. A GUI client provides a much
richer experience. From a scalability point of view the reduced
bandwidth requirement and local processing on the client has got to be
a good thing. With caching on the client even disconnected use is a
possibility.
My fiddling has been based on a client-server mode; alturistically, a
peer-to-peer solution would be much better. A client-server model seems
the easiest to implement, but relies on someone providing a server for
all of time. Should it take off, this server would have high bandwidth
costs, etc. so leans towards a commercial subscription requirement.
This in turn limits the number of casual users who may take an
interest. A peer-to-peer solution initially seems rather complicated,
but has many benefits if it were to work (all inclusive user-wise,
worries with copyrighted material, e.g. census images are passed off
elsewhere, scalability, resilience).
Since your posting, I've read Hilton Campbell's Distributed Family Tree
blog at http://blog.nucleartoiletpaper.com/dft/. This looks like quite
a neat way to go about things. There's some work been done of
distributed RDF (RDFPeers) although I don't think it quite suits, it is
a good starting point. I'm not yet convinced RDF is a great deal better
than a big well-structured distributed database (could well be less
efficient) but it makes for easier building-block layers.
> To get acceptance from the users the program must hide the complexity of
> Gentech in most cases. It should only be visable in 'expert mode'. So we
> need a mapping between the simple 'traditional' model ("My grandmother's
> name was ... and she was born on ...") and Gentech's evidence/assertion
> model.
Hiding the complexities of the GDM from the casual user is a difficult
problem, but one that also allows opportunities. For example, if the
multiple-persona concept can be handled, it makes it easier to merge
data from multiple users on the same person.
--
Mat