[Serdev] Group membership functions

[Jan Janak]

On 24-07 21:16, Jiri Kuthan wrote:
> At 12:19 PM 7/24/2003, Jan Janak wrote:
> >  I wouldn't proceed with a general caching mechanism here, caching can
> >  be highly optimized for a specific task and that will be faster.
> 
> I don't entirely agree. Primary improvement is achieved by using cache,
> cache and hashing strategies are imho secondary and may be parametric
> to accomodate specific tables. ReImplementing cache mechanism for each 
> database sounds like a headache to me.

 A generic database caching should be performed by the database client
 library, not by the application, many of the libraries probably do it anyway 
 because  very often the data is tranfered over the network which is time 
 consuming.

 The modules typically need to convert the data retrieved from the
 database to some form which can be directly used. This should be done
 once and the cache should be designed for the converted data.

 The cache should be as close to the data structures that the modules
 need as possible (in other words, no conversions should be done when
 using the cache).

 Creating a good generic database cache that would be able to satisfy 
 different needs of different modules could IMHO be more complicated.

 my 2 cents, Jan.