Hi people!

Well, I jump in again as the eternal peace keeper - we Swedes are
generally good at that job in fact. :)

And you americans suck on it. ;-) :) (<-- extra smiley so that noone get
upset, it was just a joke)

Colin Putney <cputney at wiresong.ca> wrote:
> On Jul 10, 2004, at 7:17 PM, Ramiro Diaz Trepat wrote:
[SNIP of Ramiro's post that... well, it came on pretty hard too like
gazoline on fire :)]

First I want to defend Avi here - he is a friendly soul :) and I have
never seen or heard him hurt a fly. Really. I still think Avi
overreacted though, but that happens to anyone. :)

> Well, it looks like this thread is just about finished. That's a shame, 
> because for a moment there I thought some progress might come out of 
> it. I didn't detect any lack of respect in Avi's posts. He spent 
> several hours testing Magma, with, as far as I can tell, the best of 
> intentions. The result? Condescending dismissals of his feedback and 
> admonishments not to rock the boat.

Now Colin is being a bit harsh too - guys, you need to cool down. :) I
am not sure I read "condescending dismissals" - in fact, it looked to me
that Chris did a whole lot of work to follow each step and try to
replicate it and find out what it was.

Sure, Chris should also eat a bit of "humble pie" (or whatever that
saying goes) because Chris - you do need to be a bit more humble and
acknowledge the fact that there *may* be issues. And this includes
documentation misses etc. A user is never "wrong" you know. :)

I am not sure what "admonishments not to rock the boat" means or refers
to.

[SNIP of Colin's good wisdom regarding feedback]

> First, Chris, you have to understand that Magma is not as stable as you 
> think it is. When I tried it out (a while ago), excited at the prospect 
> of a Squeak-native OODB, I ran into strange problems like the ones Avi 
> describes. Unlike Avi though, I just shrugged, said "this doesn't 
> work," and moved on. Your response to Avi and others who report 
> problems has been to deny that there is a problem, which is a signal to 
> potential users that you don't intend to fix it, and that reporting 
> bugs is pointless. I know this is probably not your intent, but it is 
> the subtle message you are sending when you respond to criticism.

Yeah, in short Colin is right here. :) And yes, we all know some of the
problems are due to "false expectations" that in themselves are of
course due to lackings in the docs - but we will of course fix that as
we go.

For example - the transaction boundary thing is very natural to Chris
and me - but not to others. But the fact that this only covers "held
objects" and not newly fetched ones is typically something that would
make me wonder too (in GemStone the isolation works for newly faulted
objects too).

> Second, I'm a little puzzled by your non-support of concurrent sessions 
> in a single image. This is precisely the configuration that would make 
> Magma useful for web-applications. Web applications developers are 
> probably the single largest group of people interested using an OODB 
> from Squeak. They're also vast majority of those using Squeak for 
> production software as opposed to research. They are the core of your 
> target market. (Again, assuming your goal is to have people use Magma.) 
> Why do you ignore them?

My guess is that the architecture of Magma (with each session having
their own set of objects) has simply not been adapted for it. If you
read Chris post he asks very clearly how Avi thinks this should be done
- he is asking for advice. So I wouldn't say Chris is "ignoring" them.

Sidenote: My web app uses a single Magma session with web requests going
through a mutex. Sure, I have no concurrency but I still have a working
persistence mechanism for my domain model.

> Finally, you should know that you come across as being rather cavalier 
> about risks to people's data. In a business setting, the kind of data 
> that goes into a database is often absolutely crucial to the business. 
> There can be no possibility of losing it. If you appear to be less 
> paranoid about data loss than your users, that in its self is a reason 
> to avoid Magma, regardless of its technical merits.

Again, the transaction log is on the todo list. We all know it isn't
there and that it is needed for true trustworthiness. This does NOT mean
you can't use Magma for stuff! Everyone aren't building mission critical
applications! You just need to be aware that pulling the plug on Magma
is NOT good. So you need to be aware of that fact and either take the
risk or have some kind of backup involved. Simple as that.

> Respectfully,
> 
> Colin

Ok, Colin's post is a good post, even if he did push it a bit with that
"condescending part" IMHO. :)

Now people - PLEASE. Cool down, and get back to positive discussion. I
would LOVE to hear from Chris at this point saying that he wants Avi's
thoughts and feedback - because I know he does. And I would LOVE for
this feedback etc to continue on a positive note.

The written word is so hard sometimes - I think you are just
"incompatible in style" as some of us unfortunately are. I got into one
of those threads a while back and they are so energy consuming - and for
what? A bloody misunderstanding? Bah.

cheer up and keep on Squeaking, regards Göran