On Wed, Mar 28, 2007 at 04:48:49PM -0500, Scott T. Hildreth wrote:
>
> > p.s. Any volunteers to work on a daemon wrapper for DBI::Gofer::Execute
> > so it can be used as a replacement for the current DBI proxy server.
> > (DBD::Gofer is potentially significantly faster than DBD::Proxy.)
>
> Are you envisioning DBD::Gofer replacing DBD::Proxy?
For many but not all possible uses, yes.
Two advantages DBD::Proxy has over DBD::Gofer are support for
transactions and very large result sets.
(Can anyone thing of others?)
DBD::Gofer may gain support for both but probably not any time soon.
Patches welcome, of course :)
Um, another is that DBD::Proxy is probably slightly faster for an
individual method call, but DBD::Gofer generally works out cheaper
because it'll make fewer round trips.
> If so, would a daemon process enable transactions for DBD::Gofer?
> Probably still stateless, thought I would ask.
I can see two ways to approach transactions for DBD::Gofer:
- bundle up a series of DBI calls into a single request that the
server then executes as a transaction and commits at the end.
- Keep a connection open between gofer client and gofer server
and loose the 'stateless' nature of Gofer for the duraction of
the transaction.
Both have pros and cons, both could be supported.
Tim.
