|Subject||Re: [IBDI] Firebird 1|
On 2001.06.01 10:50:03 -0400 Ann W. Harrison wrote:
> At 01:38 PM 6/1/2001 -0700, Peter Morris wrote:
> >So I should hold a dataset open on the server just to allow this ?
> >Surely that is taking up more resources than I need !
> Holding a dataset open on the server is much cheaper than
> reopening it.
isn't this a time-space trade off?
holding open on server -- holds space on server, next execution very fast
---perhaps most appropriate when there are few users, connected for a long
closing and reopening -- frees space on server, reallocated slowly on next
--perhaps appropriate when there are zillions of users connected for a
short time, most of whom don't come back to reexecute.
I think maybe we are asking the wrong questions here. Most of the "I want
limit n, m no matter how slow" arguments seem to be based on the idea of
using a rdms for a lousy web search engine, wherein you search for
something and almost all the numerous results are totally irrelevant, so
you have to look through pages of goo to find out you asked the wrong
question. Perhaps a transactional resource is not the appropriate
repository for queries of this nature. In such a situation, it hardly
seems like you can guarantee ever showing the most appropriate result. How
do web search engines work? Perhaps periodic extractions of data and using
something like a web search engine would be more appropriate for
applications of this kind.
I think a better although perhaps harder question is, how can we show the
user the 7+-2 items they are both interested in and can pay attention to at
the same time. Then the result set is manageable and we can just return
all the rows.
> We have answers.
> Community email addresses:
> Post message: IBDI@yahoogroups.com
> Subscribe: IBDIemail@example.com
> Unsubscribe: IBDIfirstname.lastname@example.org
> List owner: IBDIemail@example.com
> Shortcut URL to this page:
> Your use of Yahoo! Groups is subject to http://docs.yahoo.com/info/terms/