Re: Broken SSL tests in master

On Thu, Dec 1, 2016 at 9:31 PM, Michael Paquier
<michael(dot)paquier(at)gmail(dot)com> wrote:
> On Thu, Dec 01, 2016 at 03:17:34PM -0500, Robert Haas wrote:
>> It might be that (as suggested downthread) we should consider
>> supporting multiple IPs in the hostaddr string as well, but that
>> requires some thought. For example, what happens if, for example, the
>> host and hostaddr lists are of unequal length? Would we accept one
>> host and >1 hostaddrs? Probably makes sense to just apply the host to
>> every hostaddr. >1 host and 1 hostaddr? Probably doesn't make sense,
>> but I guess you could argue for it. Equal length lists definitely
>> make sense.
>
> That would make the current code a huge plate of spagetthi for sanity
> checks considering the multiple interations between port, host and
> hostaddr. It seems to me that the current approach of supporting only
> port and host is simple enough and will satisfy most of the user's
> need plently. So +1 for simplicity.

I don't think it'd be ridiculously complicated to make it work and I
don't mind if someone wants to try. However, it wasn't interesting to
me, so I didn't spend time on it. A lot of these parameters are
intertwined, and I wanted to avoid trying to boil the ocean. But I'm
not allergic to follow-on patches.

--
Robert Haas
EnterpriseDB: http://www.enterprisedb.com
The Enterprise PostgreSQL Company