On September 30, 2015 at 7:06:58 AM, Tom Lane (t...@sss.pgh.pa.us) wrote: Paul Ramsey <pram...@cleverelephant.ca> writes: > Hm. Wouldn't it be just fine if only the server is able to define a > list of extensions then? It seems to me that all the use-cases of this > feature require to have a list of extensions defined per server, and > not per fdw type. This would remove a level of complexity in your > patch without impacting the feature usability as well. I would > personally go without it but I am fine to let a committer (Tom?) put a > final judgement stamp on this matter. Thoughts?
Maybe I'm missing something, but I had envisioned the set of safe-to-transmit extensions as typically being defined at the foreign-server level. The reason being that you are really declaring two things: one is that the extension's operations are reproducible remotely, and the other is that the extension is in fact installed on this particular remote server. Perhaps there are use-cases for specifying it as an FDW option or per-table option, but per-server option seems by far the most plausible case. Fair enough. Declaring it for the whole database as an option to CREATE FOREIGN DATA WRAPPER was just a convenience really, so you could basically say “I expect this extension on all my servers”. But you’re right, logically “having the extension” is an attribute of the servers, so restricting it to the server definitions only has a nice simple logic to it. P. -- http://postgis.net http://cleverelephant.ca
