2015-02-27 19:32 GMT+01:00 Stephen Frost <sfr...@snowman.net>: > * Pavel Stehule (pavel.steh...@gmail.com) wrote: > > 2015-02-27 17:59 GMT+01:00 Stephen Frost <sfr...@snowman.net>: > > > I don't think we actually care what the "current contents" are from the > > > backend's point of view- after all, when does an individual backend > ever > > > use the contents of pg_hba.conf after it's up and running? What would > > > make sense, to me at least, would be: > > > > > > pg_hba_configured() -- spits back whatever the config file has > > > pg_hba_active() -- shows what the postmaster is using currently > > > > I disagree and I dislike this direction. It is looks like over > engineering. > > > > * load every time is wrong, because you will see possibly not active > data. > > That's the point of the two functions- one to give you what a reload > *now* would, and one to see what's currently active. >
> > * ignore reload is a attack to mental health of our users. > > Huh? > this topic should be divided, please. One part - functions for loading pg_hba and translating to some table. Can be two, can be one with one parameter. It will be used probably by advanced user, and I am able to accept it like you or Tomas proposed. Second part is the content of view pg_hba_conf. It should be only one and should to view a active content. I mean a content that is actively used - when other session is started. I am strongly against the behave, when I have to close session to refresh a content of this view (after reload) - and I am against to see there not active content. Regards Pavel > > > It should to work like "pg_settings". I need to see "what is wrong in > this > > moment" in pg_hba.conf, not what was or what will be wrong. > > That's what pg_hba_active() would be from above, yes. > > > We can load any config files via admin contrib module - so there is not > > necessary repeat same functionality > > That's hardly the same- I can't (easily, anyway) join the results of > pg_read() to pg_hba_active() and see what's different or the same. > > Thanks, > > Stephen >
