[mapguide-internals] RE: RFC 86 some thoughts (Zac Spitzer)

[Leaf Li]

Hi Zac,

Thanks for your feedback. Comment in lines.

i think a limit of pinned connections per fdo provider would be good
so you can not exhaust the number of available connections thru
pinned ones

"There are only three methods supporting transaction, ..."

explaining which ones aren't help paint the picture a bit

in a typical web app, the session which your example assumes
which is the expected norm for web developers, is that you would
use a transaction for

as such most web apps tend to only ever hold a transaction
for a request

[Leaf] It isn't true. I use transaction in this example is just to show intention.
Actually you can take a look at the second example. Currently two reader will
results in two FDO connection are occupied because server doesn't return
results in one time. If using the pinned connection, we can reduce FDO connections used
by client to 1.

consider the example

log in, pin an oracle connection, update one row, do nothing

would that create a db lock unless something else is done for 20 min?

sessions != requests

might need some sort of purge all pinned connections, as short
of a server restart, 20 mins is an outtage

[Leaf] Yes. It is true that connection will occupied by the current session
if users don't call UnpinConnection(). However, if so, it is a programming
error. Typically, the pinned connection is used in a short time. The usage of
the pinned connection is similar to feature reader. For feature reader, users
need to close feature reader once they don't need it. Users need unpin 
connection too once they don't need it.

However, I don't object to have a time out setting for the pinned
connection.

Thanks,
Leaf Li