sqlite file format, anyone?

[Derek Atkins]

Matthew Vanecek <mevanecek at yahoo.com> writes:

> > Except it's not very real-time (the client has to emit a query
> > to notice the event), and it's not very portable (it requires
> > pg).
> 
> The libpq library is documented to be thread-safe, so I could put the
> notify polling in a fork(), using select() or pselect(), and have it
> generate a signal when it receives a notification.  Pretty simple to
> accomplish in theory, but would require some signal handling
> architecture.  Doesn't take up any extra CPU cycles, either.

Well, I think I'd rather use pthreads rather than a "fork()", but
sure, that works for me.

> The SQL back end should be inherently multi-user.  There's a certain
> amount of checking that can theoretically be safely skipped in a single
> user environment, but that's skipping I plan on adding later.  For the
> moment, it's a multi-user architecture.  For me, this is mainly a
> difference at the very lowest levels.  The upper layers shouldn't care
> what type of access is being used (other that storing it in the Backend
> object).

Ok, this implies we cannot do multi-user on MySQL or (theoretically)
SQLite, and also implies we can't make a general libdbi interface
because we wouldn't be able to handle the notify events.

That's fine with me so long as we're clear on the ramifications.

-derek
-- 
       Derek Atkins, SB '93 MIT EE, SM '95 MIT Media Laboratory
       Member, MIT Student Information Processing Board  (SIPB)
       URL: http://web.mit.edu/warlord/    PP-ASEL-IA     N1NWH
       warlord at MIT.EDU                        PGP key available