[Schevo-devel] Re: concurrency and copy-on-write snapshots

[Matthew Scott]

On Jun 22, 2006, at 8:35 AM, Jofell Gallardo wrote:

> Hi guys... I would want to contribute to your project. It's my  
> dream to get involved with an object db project (better if open  
> source), but I dunno how to start. i see your feature-lists but  
> dunno how the these things work, actually :P....

Thanks for joining :)

Right now, Schevo is under-documented, and we are all too aware of  
that.  :(  However we do have some basic documentation on the Schevo  
Trac that could help get you started, and we are glad to assist you  
here on this list (and in the #schevo channel on freenode).

One thing to keep in mind about Schevo is that it differs from other  
object databases in some ways.  Schevo restricts the structure and  
API of the database schema and the data itself.  This restriction  
affords lots of automation as far as indexing and keeping track of  
object relationships.  It also provides a nice, coherent API that  
once you are accustomed to, works in your favor to promote rapid  
development.

In many ways, you can translate RDBMS concepts more or less directly  
to Schevo concepts as far as structuring your data.  This may be  
useful to help get started, and as you become more familiar with  
Schevo you'll find ways in which Schevo extends beyond what you're  
accustomed to.


> I do know python and other languages, and worked with BDB before  
> (but not that hardcore)...
>
> Also, I guess it's a good suggestion that we employ BDB for Schevo,  
> to use it's DB replication feats and stuff... if ever I may be  
> missing some  points, please enlighten me.

If the BDB backend for Durus that Jesus Cea is working on is  
compatible with how Schevo uses Durus, I can see that as being a  
definite option once we use the latest version of Durus (see http:// 
schevo.org/trac/ticket/5)

As far as DB replication, I am not so sure that read/write  
replication would be possible simply by using BDB, since there are  
synchronization issues that are probably best left handled at the  
Schevo level.

In Schevo 2 (which predates the use of Durus in Schevo), we had a  
simple replication service available that allowed multiple clients to  
have a local database synchronized with the server.  It was used for  
development, not in any live deployments.

We don't yet have something like that in Schevo 3, simply because we  
haven't found a pressing need for it yet, and Schevo is very much  
developed on the basis of real-world use cases and needs.  I'm sure  
the replication need will become more urgent in the coming months,  
though. :)

More near term will be the ability to interact with a Schevo database  
server using HTTP.  That capability is being developed in a separate  
project (open source-licensed, to be made available in July) but may  
be merged into Schevo itself at some point once it matures.


--
Matthew Scott
mscott at springfieldtech.com