Greg Smith wrote:
> On 04/03/2011 02:18 PM, Joshua Berkus wrote:
> > * Transaction-controlled Synch Rep
> > I would like to have volunteers from the advocacy list commit to taking on one of each of these features. For each
onewe need:
> >
> > a) a two-line explanation of what the feature is and why it's valuable (for the release notes, etc.)
> > b) a wiki page with a more detailed explaination and examples oriented towards the beginning-to-intermediate
PostgreSQLuser.
> >
>
> We already have http://wiki.postgresql.org/wiki/Synchronous_replication
> for sync rep, and I just started changing that around so that it
> reflects the code committed into 9.1. I got my first set of questions
> today suggesting our internal work on documenting this from the ground
> up is moving along. Over the next month we'll have at least two people
> chugging away at making that targeted more toward beginners (or split
> into something that is).
>
> I've been doing the elevator pitch for sync rep for a while now; here's
> a first draft description for the release notes:
>
> Transaction-controlled Synchronous Replication: When replicating to
> multiple nodes, customize every database transaction for its individual
> speed and durability needs. Options range from only committing to
> memory on the master up to the new synchronous standby mode, where data
> must be stored on multiple servers to be considered safe.
...
> PostgreSQL 9.1 lets you pick exactly the level of commit guarantee your
> data requires. Whether you want unlogged tables optimized only for
> speed, or you need durable synchronous replication to multiple servers,
> you're covered--all in one database.
I think the "users get control" montra could be a theme for this
release; I think it would apply to all of these:
* Unlogged tables to better handle some NoSQL workloads
* Synchronous replication for greater reliability
* SQL/MED (Management of External Data) (flat files, other databases)
* Per-column collation support
* Security Label, including SE-Linux integration
* True serializable isolation with predicate locking (already had snapshot isolation)
> Need to take a shower to wash the stench of marketing off now.
LOL
--
Bruce Momjian <bruce@momjian.us> http://momjian.us
EnterpriseDB http://enterprisedb.com
+ It's impossible for everything to be true. +
