Re: Updates of SE-PostgreSQL 8.4devel patches (r1268) - Mailing list pgsql-hackers

From KaiGai Kohei
Subject Re: Updates of SE-PostgreSQL 8.4devel patches (r1268)
Date
Msg-id 493FB75A.2020203@kaigai.gr.jp
Whole thread Raw
In response to Re: Updates of SE-PostgreSQL 8.4devel patches (r1268)  (Greg Stark <greg.stark@enterprisedb.com>)
List pgsql-hackers
Greg Stark wrote:
> Is there any reason it's easier to do as a configure option instead of 
> an initdb option or better yet a per-database option?

SE-PostgreSQL needs "libselinux" to communicate with SELinux,
however, it is not available for non-SELinux'ed platforms.

> The reason we're shying away from configure options for functionality 
> changes is that more and more users are getting pistgres from binary 
> distributions. Which option should redhat ship for example?

As we're doing on Fedora, SE-PostgreSQL will be distributed as
a separated package.
If a user want to start SE- version, he will install another
binary package, and stop vanilla PostgreSQl, and start SE-PostgreSQL.

> Also, a configure option is kind if weird since that's a property of the 
> binary not the data: what happens if you build a database with selinux 
> and the switch binaries to a non-we build? Is that option in pg_control?

In this case, the vanilla PostgreSQL should be simply ignore the
data which is generated by SE-PostgreSQL.

Thanks,
-- 
KaiGai Kohei <kaigai@kaigai.gr.jp>


pgsql-hackers by date:

Previous
From: Bruce Momjian
Date:
Subject: Re: WIP: default values for function parameters
Next
From: Heikki Linnakangas
Date:
Subject: Re: Sync Rep: First Thoughts on Code