On 6 February 2015 at 16:13, Bert Huijben <b...@qqmail.nl> wrote: > > >> -----Original Message----- >> From: Ivan Zhakov [mailto:i...@visualsvn.com] >> Sent: vrijdag 6 februari 2015 14:04 >> To: Bert Huijben >> Cc: Subversion Development >> Subject: Re: svn commit: r1657782 - /subversion/branches/reuse-ra- >> session/BRANCH-README >> >> On 6 February 2015 at 15:55, Bert Huijben <b...@qqmail.nl> wrote: >> > >> > >> >> -----Original Message----- >> >> From: br...@apache.org [mailto:br...@apache.org] >> >> Sent: vrijdag 6 februari 2015 12:10 >> >> To: comm...@subversion.apache.org >> >> Subject: svn commit: r1657782 - /subversion/branches/reuse-ra- >> >> session/BRANCH-README >> >> >> >> Author: brane >> >> Date: Fri Feb 6 11:09:54 2015 >> >> New Revision: 1657782 >> >> >> >> URL: http://svn.apache.org/r1657782 >> >> Log: >> >> On the reuse-ra-session branch: Documentation change: Instead of >> expecting >> >> clients to manage the lifetime of idle sessions, expect that the RA cache >> >> implementation will take care of that itself and hide the complexity >> >> from the API. >> >> >> >> * BRANCH-README: >> >> Remove todo item about svn_client_close_all_sessions. >> >> Add todo item about idle session expiry. >> > >> > How do we expire idle sessions if a client only uses Subversion api's once >> > a >> day? >> > >> > I don't think we want to rely on timers or threads. >> > >> One solution is to record last time RA session used and check it when >> we're going to reuse this particular RA session. > > That doesn't reduce the usage of limited resources to until we try to use > them later. > The limit on number of idle open RA sessions is already in TODO list: - Limit the number of idle open sessions in the cache.
I don't see problem with implementing svn_client_close_all_sessions(). My point was that Subversion should work fine (no connection timeout errors) even application never calls svn_client_close_all_sessions(). -- Ivan Zhakov
