New Account Types and data equivalency (was post TB 3.1 mailnews backend plans)
KaiRo at KaiRo.at
Tue Jul 20 14:38:04 UTC 2010
Andrew Sutherland schrieb:
> I think you may be making some assumptions about the implementation and
> design decisions that are not consistent with my plans. It's also worth
> noting that even if we were to store all authoritative data exclusively
> in SQLite with no export support, the SQLite database format is well
> documented ( http://www.sqlite.org/fileformat.html ), the source is in
> the public domain and it compiles on basically every platform available.
> We would have to go really far out of our way to create an equivalent of
> the pst situation back in the day.
The only positive thing of pst is that every single bit of info of an
installation can be backuped by one file, and we will never get there.
On the other hand, single large files, esp. with the way gloda is
exploding already right now, run into the problem of being destroyed
much more easily in on-disk data corruptions, be it from filesystem or
hardware causes (and the latter become increasingly likely with the rise
of flash-based memory) - and that even assumes that the data format
doesn't get corrupted internally.
There's also a reason why places creates very frequent backups of its
bookmarks data, as it has not been proven to be as reliable as one would
hope for important data, and most people consider their messaging data
to be even more important than their bookmarks, so we need to have a
solution for this.
Also, I seem to remember of hearing advice that throwing away gloda and
letting Thunderbird rebuild it helps with some problems people are
seeing, so I'd hope we'll have a mechanism to still do that. If it means
we do backups of the really important data in JSON format during idle
times, like places does for bookmarks, and restore that when gloda is
found empty or corrupted, that's OK. I just don't want us to end up
causing profiles to be more unstable as they are right now.
More information about the tb-planning