TDB2 Alpha release....

Rusty Russell rusty at rustcorp.com.au
Wed Dec 8 05:35:34 MST 2010


On Thu, 2 Dec 2010 02:40:34 am simo wrote:
> On Thu, 2010-12-02 at 00:04 +1030, Rusty Russell wrote:
> > Hi all!
> > 
> >     After a few weeks of hacking, TDB2 is now almost presentable.  There's
> > stuff missing (both in terms of minor tdb1 functionality and the tdb2 design
> > doc) but the basic format is now working, and transactions are implemented.
> > 
> > I'll be hacking more on it in the coming days, but you can find the current
> > state (as always) in CCAN, as a tarball or in the CCAN git tree:
> > 
> > See:
> > 	http://ccan.ozlabs.org/info/tdb2.html
> > 
> > In particular, the design doc has been updated with status.
> > 
> > My next immediate tasks are:
> > 1) Implement missing functionality from tdb1.
> > 2) Implement the promises in the design doc re: extensibility and
> >    multiple opens.
> > 3) Performance testing using ldb traces.
> 
> Hi rusty, id tdb2 supposed to be API compatible ?

No, there were too many things we wanted to clean up.  A wrapper would
be possible though (for your point below).

> (I guess not).
> 
> If not wouldn't it make sense to rename functions to use the prefix
> tdb2_ so that if you combine 2 libraries that use the 2 tdb versions at
> the same time you don't get nasty name conflicts ? (It would also allow
> slow adoption within the samba source code).

One option was to put the whole tdb1 source code into tdb2 as a fallback
for old tdbs.  Then it's just a question of API changes; we can bump the
soname major number.

Maybe it's less painful to live with the "tdb2" prefix, even though in
five years it's going to look pretty ugly?

I don't really have a strong opinion...
Rusty.


More information about the samba-technical mailing list