Subject | Re: JayBird 1.5 features |
---|---|
Author | Roman Rokytskyy |
Post date | 2004-03-01T23:11:12Z |
Hi,
WHERE CURRENT OF myCursor, then it is already supported in 1.0.1. If
you mean updatable result sets, no this is not planned for 1.5
(unless somebody contributes tested code :)
same host. In case of type 2 JDBC driver it will use IPC instead of
sockets, in case of embedded engine driver it will access database
file directly. Usually this gives approx. 2x time performance boost.
If you access server via network sockets, type 4 (i.e. pure Java)
driver is fastest (type 2 in my tests was ~10% slower).
situation where type 4 driver might die, is when core engine team
decides to completely change network protocol and will make code more
obfuscated than it is now. However I hope that in case of wire
protocol change all modifications will be documented. Anyway, there
were no change in 1.5, and currently no change is planned for 2.0.
There were talks in FB-Architect group, but as far as I know, that's
still talks. So, it seems that for the nearest 1-2 years there will
be pure Java driver.
Best regards,
Roman Rokytskyy
> 1. Will it support updateable resultsets ?If you mean Statement.setCursorName(...) and following UPDATE ...
WHERE CURRENT OF myCursor, then it is already supported in 1.0.1. If
you mean updatable result sets, no this is not planned for 1.5
(unless somebody contributes tested code :)
> 2. Why does it use (dll/so)s ? Will it be faster that way ? WhatType 2/embedded version makes sense when you access database on the
> about stability [especially in classloader contexts where the
> driver may be instantiated multiple times] ?
same host. In case of type 2 JDBC driver it will use IPC instead of
sockets, in case of embedded engine driver it will access database
file directly. Usually this gives approx. 2x time performance boost.
If you access server via network sockets, type 4 (i.e. pure Java)
driver is fastest (type 2 in my tests was ~10% slower).
> 3. Will it also still be available in a pure-Java version?Sure. We do not plan to sacrifice type 4 driver. The only possible
situation where type 4 driver might die, is when core engine team
decides to completely change network protocol and will make code more
obfuscated than it is now. However I hope that in case of wire
protocol change all modifications will be documented. Anyway, there
were no change in 1.5, and currently no change is planned for 2.0.
There were talks in FB-Architect group, but as far as I know, that's
still talks. So, it seems that for the nearest 1-2 years there will
be pure Java driver.
Best regards,
Roman Rokytskyy