https://bz.apache.org/bugzilla/show_bug.cgi?id=58590
--- Comment #2 from Aidan <apa...@leocap.com> --- (In reply to Konstantin Kolinko from comment #1) Hi Konstantin, > 1. Code conventions: > http://tomcat.apache.org/getinvolved.html > > Actually Sun's conventions but with spaces instead of tabs. Ok, I can edit to conform with that. > 2. backgroundProcess() runs frequently. I see no need to perform this work > on each run. Every 10 seconds for this container, by default, I believe - that doesn't seem too often to merely check the modified date of a file. > 3. It shall work without stopping and starting the realm. Users should not > be locked from the system while it reloads. OK, fair point, I can try to address that. > 4. MemoryRealm is rarely used. The usual configuration uses > UserDatabaseRealm + MemoryUserDatabase (created by > MemoryUserDatabaseFactory). True, but it wasn't clear to me how to effect the same change in UserDatabaseRealm and I was a bit pushed for time. > 5. It should be possible to turn this feature on or off. As I wasn't proposing this change for the default Realm I didn't think that was necessary. > I do not know what the default should be. > On one hand we already have <Host autoDeploy="true"> so we are already > checking hard drive by default. > > On other hand, on production systems such setting is likely to be off by > default (as there is an expectation that nobody will ever update that file). I think that setting only relates to .war files ?? AFAIK there is currently no mechanism to force re-reading of tomcat-users.xml without a restart. But I accept that if a user requires runtime user/role loading then they should probably use a JNDI or JDBC-based realm instead. However, I think it's nice to have an out-of-the-box alternative that has this capability. I wrote this patch at the bank where I work where we use Tomcat (with the APR) to serve several hundred internal users 24x5. We rarely update tomcat-users.xml but when we do we have to perform a disruptive restart. The patch fixes that, at least. > 6. There shall be an explicit method to reload and an explicit method to > perform an up-to-date check - so that it were possible to call them via JMX. I hadn't thought of that. But as cool and useful as JMX is, it's beyond a lot of the casual users whom this patch is aimed at, I suspect. Having said all that, I'm happy to rework it as best I can in order to address all your points or for you to simply reject it. Kind Regards, Aidan -- You are receiving this mail because: You are the assignee for the bug. --------------------------------------------------------------------- To unsubscribe, e-mail: dev-unsubscr...@tomcat.apache.org For additional commands, e-mail: dev-h...@tomcat.apache.org
