Toby DiPasquale on 30 Dec 2005 20:20:15 -0000

[Date Prev] [Date Next] [Thread Prev] [Thread Next] [Date Index] [Thread Index]

Re: [PhillyOnRails] Re: Is anyone using Object Prevalence with Rails?

On Fri, Dec 30, 2005 at 12:42:50PM -0500, Eric Snyder wrote:
> I personally don't have any experience using Object Prevalence, however 
> I've read several posts similar to these 
> <> discussing the use of 
> Madeline with Instiki. Not having any experience with Instiki or Madeline 
> I can only speculate on the cause of the instability, it may be an 
> inappropriate use of Object Prevalence and not a problem with the Madeline 
> implementation itself. 

Yeah, I saw this one. I'm glad you brought this up. I had said in my last
email that I thought my idea was original but then found out that it
wasn't. Well, turns out that it still is, at least part of it. OP systems
assume that the app will always be on and that data can fit in RAM. The
second is a better assumption than it used to be, but the first is just
plain ridiculous. Only Erlang programmers enjoy that sort of uptime ;-)

Seriously, though, I was thinking of something more along the lines of an
intelligent object caching system. OP is a little less sophisticated than
what I was referring to, but the basic concept is definitely what I was
looking for. To me, relational databases don't really fit with the current
model of application data. SQL is good for ad-hoc queries, something that
almost no Web application does. As well, mapping your OO logic (or even
imperative) onto the relational model forces your entire application to
think in relational set theory, which frankly, I couldn't care less about
since it solves problems that I don't have.

Expect to see some software from me in the future addressing this issue.
It may not map directly onto ActiveRecord, but I hold more promise for CPS
frameworks, anyway ;-)

> I did find it interesting that Madeline does not currently have the ability 
> to roll back commands and that commands are responsible for ensuring that 
> changes will be successful prior those changes being made. Without the 
> ability to rollback, I assume that commands are responsible for dealing 
> with exceptions and must themselves be able to guarantee that execution 
> cannot put the system in a bad state.

Yeah, keeping the system in a consistent state is the purview of the
commands in an OP system, it would seem. I think of them as "stored
procedures for object-based datastores". They regulate access to the data
and keep the system humming along as many letters of the ACID acronym as
you want (assuming your implementation supports them, that is). The
"commands" are necessary from an atomicity and consistency perspective.

> BTW: Excellent presentation at the last meeting.


> >[4]

P.S. This one appears to be vaporware thus far. Sorry.

Toby DiPasquale
talk mailing list