File store (was Re: Salutations)

[Derek Atkins]

David Merrill <dmerrill@lupercalia.net> writes:

[snip]

> But I do take issue with the argument that period closing, and the
> transfer of the account balance, is the way to accomplish this. It is
> certainly one way, but there are others. Let me propose one other way
> for your consideration.
> 
> Let's say we have a table that looks like:
> 
> account_num
> date
> closing_balance
> 
> Each record represents the end-of-the-day balance for the given
> account on the given date, which means the balance after all
> transactions for that day have been calculated.
> 
[snip]
>
> Does this accomplish the same thing you were proposing? Am I
> understanding your intent now?

Unfortunately I'm not convinced it completely solves the problem.  The
problem with any checkpointing is that if any transaction changes
before the checkpoint, then you have to modify all checkpoints that
have occurred after the transaction.

The nice side-effect of "closing the books" is that theoretically all
transactions dated prior to the closing are considered 'immutable.'
Perhaps this is just a different type of checkpoint?  I'm not sure.

If we do have the ability to make transactions prior to a particular
date 'immutable', then we only have to worry about checkpoint
consistency for dates after that date.  Perhaps we need the notion of
'closing the books', whereby transactions become immutable, and
'checkpointing', where transactions are still mutable so the
checkpoints may change as transactions get updated.

-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@MIT.EDU                        PGP key available