Developers of Sympa

[Sergiy Zhuk <address@concealed>]

hi

On Fri, 20 Sep 2002, Aumont - Comite Reseaux des Universites wrote:

> List configuration are not the only data that servers share when running 
> Sympa services on  multiple machines : archive, templates, scenario and 
> SPOOL also are needed.

True, but having configuration in a database also provides a nice interface
to sympa configs for third party software and allows to update some things
like list owners and editors in a more efficient and simple manner than now.
It also makes mass config changes much simpler.
In addition, a lot of what sympa is doing can be optimized by using more
selective queries to the database rather than keeping all list info in
memory cache, which takes about 600kb of RAM per list from my experience.

I understand the problems with changing the way configs are handled though.
So, I think a common cache could solve some of the issues we're having w/o
changing the config storage type for now and changing a lot of code in
general.
Maybe we could start with implementing a common cache first for the
following reasons:
- performance will improve, since we could use a dedicated process to update
  it, i.e. other processes will assume information is current and won't
  access the configs on disk at all
- memory consumption will decrease
- we don't have to change config storage right away, i.e. replacing current
  cache code is probably simpler than re-designing config storage and
  optimizing sympa code to use the advantages of dbase storage
- we can provide an API to that common cache to allow third party sw to
  update the cache directly for list configs updates and such w/o knowing an
  underlying storage structure

thanks

-- 
rgds,
serge