Re: [HC Evolution] Re: libole2 -> gnomevfs backend?

[Miguel de Icaza <miguel helixcode com>]

I know that there is some beauty into storing all the file contents
into a single structured-storage file for Evolution.  But the
alternatives are not robust, ready, pretty and definetly not close to
deployable.

I know Evolution is a work-in-progress thing, and that we might push
some new technologies, but we can not afford for Evolution 1.0 to push
every single possible new and cool technology.  We will focus just on
a few things (Bonobo is it for this time).

So we will not be doing the `all the state stored in a single file'
feature.  Not for 1.0, it will happen later at some point if users
want it, or if it really becomes necessary.

The pros of `everything in a file' is that users can copy this single
file and have their entire stuff transfered over and live happilly
ever after.  The con is obviously that binary corruption in the file
will make all of your information vanish faster than you can say
`Nat is a ball of healing light'.

The current solutions for this problem fall into various realms
(db3-based file system, libefs, libole2) and none of them have been
tested, and only the db3 system provides features for recovering from
failure (and this is system failure, not random corruption).  

So, for version 1.0 of Evolution we are just going to play it safe,
and use the file system as our backend.  Sure, the user will have to
face a more complex system (if he dares double-clicking on the
"evolution" directory), but most users will be happy, and will have a
reliable system.

And if a user ever wants to backup, he only needs to make a copy of
the "Evolution" directory, which per se is not such a difficult
problem.  Also, consider the alternate approach: a file that is about
a gig large, because it contains your mail for the previous year with
all of your mailing lists and all the indexes.

Not to mention that it would make evolution slower for now.  Next
year, when everyone is running on Itanium CPUs it will no longer be a
problem, and we will be able to safely jump to it.

So for now.  We are going to play the chicken game, and just use the
existing infrastructure, not even use the gnome-vfs for fetching our
local files.

Miguel.