A
apandapion
I'm starting on my fourth database/web service/web client, using C# as
the language. On my last project, I had two datasets. One lived in
the Cache object, was very large, rarely changed, and was shared across
all users of the application. The other lived in the Session object,
was smaller, changed frequently, and one existed per user of the
application.
The first, the cacheobject in Cache, held all of the data that an
application uses that rarely changes - lists of users, and all of those
tables you use to populate dropdown lists. The second, the
sessionobject in Session, held the data that the user was specifically
working with. So if you have 20 users online, they share the
cacheobject and each have thier own sessionobject.
This works very well. The drop in bandwidth across all of the boxes
involved in the application is large - a similar application in old
school asp would have a copy of all of the data for each instance.
What is bothering me is that this scenario breaks relations between
tables. You have to either not add the relations between parent-child
tables or remove them if they come down from the database, because you
have half the data in one dataset and half the data in the other. Each
dataset, from it's own perspective, is missing half of the data. You
could merge the two into the sessionobject, but then you have 20 copies
of the bulky metadata, one per client.
So my questions are:
1) Can relations and constraints exist between the tables in two
different datasets? This seems unlikely, but I might be suprised.
2) If you can have constraints between two datasets, would you then
just obtain both sets of data and then apply the constraints on the web
client side?
3) Who else has considered this problem? What did you come up with?
the language. On my last project, I had two datasets. One lived in
the Cache object, was very large, rarely changed, and was shared across
all users of the application. The other lived in the Session object,
was smaller, changed frequently, and one existed per user of the
application.
The first, the cacheobject in Cache, held all of the data that an
application uses that rarely changes - lists of users, and all of those
tables you use to populate dropdown lists. The second, the
sessionobject in Session, held the data that the user was specifically
working with. So if you have 20 users online, they share the
cacheobject and each have thier own sessionobject.
This works very well. The drop in bandwidth across all of the boxes
involved in the application is large - a similar application in old
school asp would have a copy of all of the data for each instance.
What is bothering me is that this scenario breaks relations between
tables. You have to either not add the relations between parent-child
tables or remove them if they come down from the database, because you
have half the data in one dataset and half the data in the other. Each
dataset, from it's own perspective, is missing half of the data. You
could merge the two into the sessionobject, but then you have 20 copies
of the bulky metadata, one per client.
So my questions are:
1) Can relations and constraints exist between the tables in two
different datasets? This seems unlikely, but I might be suprised.
2) If you can have constraints between two datasets, would you then
just obtain both sets of data and then apply the constraints on the web
client side?
3) Who else has considered this problem? What did you come up with?