On Mar 3, 2012, at 15:54 , Mike Coolin wrote: > So don't improve the documentation such as it is, read the tests? > > I think Andrey has a point. Tests are not documentation, they are guardrails > and samples.
My point was not that we shouldn't document this. Of course we should document this, thanks for raising the issue. I took issue with these two sentences: "Problem with _purge that it was not documented. That means this was not well tested." in particular that there is somehow a causal relationship between non-documented features and their well-testedness. While it is correct that well documented features do get more real-world exposure, and thus testing, I just wanted to point out that we do have tests for _purge that have been around for a long time and that I am reasonably confident that it is a stable feature. Sorry for the confusion! :) Cheers Jan -- > Cheers Mike > > ----- Original Message ----- > From: Jan Lehnardt > Sent: 03/03/12 09:31 AM > To: [email protected] > Subject: Re: Documentation issue > > On Mar 3, 2012, at 06:57 , Andrey N wrote: > Problem with _purge that it was > not documented. That means this was not well tested. I'd contest that notion, > there's been automatic tests for _purge ever since the feature landed. Cheers > Jan -- > > Thanks. >> To me it seems exactly the same, but I'm not an expert. > >> >> On Fri, Mar 2, 2012 at 3:39 PM, Robert Newson<[email protected]> > wrote: >> >>> I said 'better' not 'good'. :) >>> >>> On 2 March 2012 20:20, > Mark Hahn<[email protected]> wrote: >>>>> A better solution is to periodically > switch to a new database and then >>>> delete the old one (when those > sessions are ended). >>>> >>>> How is that any different than purging? It > also kills replication. >
