Le 04/11/2017 à 13:28, Stefan Seelmann a écrit : > On 11/04/2017 12:50 PM, Emmanuel Lécharny wrote: >> Hi guys, >> >> >> I'm (slowly) reviewing the way we use cache in ApacheDS, as we need to >> use one in Mavibot. Currently, we have a CacheService that is managing >> the caches, hidding a bit of their initialization and such things. >> >> I'm not sure it's really well implemented : >> >> - it does not abstract the cache system we use (it's dependent on >> ehcache atm) >> >> - it just provide a way to add a cache, remove it or get it >> >> - it depends on a configuration file on disk >> >> >> The third part is really annoying, especially for those who want to >> embed the system. >> >> >> I do think we have all we need in teh ApacheDS configuration, ie the >> cache size. We don't persist anything on disk, we don't use any specific >> feature related to the cache system we use. >> >> >> Here is what I suggest we do : >> >> - make the CacheService cache agnostic : it should work with ehcahce or >> any other cache system we might want to use >> >> - mahe the Cache instance we return be an abstraction over the cache >> returned by the underlying cache system >> >> - get rid of the configuration file >> >> >> I don't think it would take long, it's quite isolated in the code (and >> that is the only good aspect of the cache service). >> >> >> If ayone of you have some better idea, or proposal, please feel free to >> express yourselves :-) > Sound good, do that change. > > The only concerns I have that > a) we are in the middle of svn->git migration and > b) the "value" branch which already lives since March 2016, which > contains changes not in trunks > > I really think we should do that first, finish the git migration and > merge back the "value" branches, so we have one clear main development > line, so other contributors are not confused.
I don't disagree :-) As I said before, I have pending changes in the value branch that I would like to commit before the commit, but considering the litle time I have to do it properly, it's probably a good timing to commit what I have, regardless of its status, do the migration, then fix the branch later. Let me do that. Thanks for the feedback, Stefan. -- Emmanuel Lecharny Symas.com directory.apache.org
