Magma....

Jimmie Houchin jhouchin at cableone.net
Sat Jul 3 16:40:01 UTC 2004


Avi Bryant wrote:
> On Jul 2, 2004, at 9:55 PM, Ramiro Diaz Trepat wrote:
> 
>> Please guys, this must have happened to someone using Magma.
>> I am using the 3.7 image and the latest available Magma.
> 
> I suspect that the reason you're not getting answers isn't that nobody 
> else has experienced the problem, but rather that nobody (except Chris) 
> knows what the solution might be, because they're all in the same boat 
> as you.  Certainly, my experience with Magma was similar to yours: I was 
> excited by the idea of a pure-Squeak OODB, but ran into errors I didn't 
> understand right off the bat, and ended up going for the easy solution 
> of finishing my GOODS client work rather than diving into the large and 
> frankly daunting Magma codebase.  I fear that this is a chicken and egg 
> problem: Magma desperately needs a few users to pound on it and submit 
> detailed bug reports and patches and get it ready for production use, 
> but won't get those users until it's ready for production use.  It's a 
> problem with the particular domain, I think: nobody wants to risk their 
> data to what's very clearly a work in progress, whereas they have no 
> problem using 80% solutions for development tools, web servers and the 
> like.  The end result is, I believe, that many people have tried Magma 
> but essentially nobody uses it.  I could be wrong, and if so, I'd be 
> very interested to hear about people's experiences with serious Magma use.
> 
> Because of this 80% solution effect, I believe that using the GOODS 
> client may actually be a more effective way to end up with a solid 
> pure-Squeak OODB solution, because it lets us tackle each half of the 
> problem separately: once we have a really good client done, we can start 
> to reimplement the GOODS server in Squeak, meanwhile falling back on the 
> C++ server for the cases where we need the reliability.
> 
> Anyway, in case it helps, what I can tell from your stack trace is this: 
> for some reason, Magma is trying to #become: one of its proxy objects 
> into a SmallInteger, which you can't do (since SmallIntegers are tagged 
> and not full objects).  The problem likely lies in whatever code created 
> the proxy in the first place (ie, the unserialization code) - you 
> shouldn't be wrapping small ints in proxies.  But that's just a guess.  
> When you say your Address object has "street and number", is number a 
> string or an integer?

I would be interested in hearing Chris' opinion on this.

As the developer of Magma and its sole maintainer, it would be 
interesting to have an idea of his vision for Magma's future. His 
commitment to that vision and ability to bring about that vision. I do 
not doubt his technical capability. But all of us have time constraints 
and other things tugging at us. Maybe even a roadmap.

Avi is being very pragmatic. The GOODS solution may in the end be the 
most expedient and practical solution. And it can be worked towards a 
pure Squeak solution over time.

Currently it seems that is easier to pick the GOODS solution. At least 
until such a time as more information is available about Magma.

I am not trying to put any undue pressure on Chris. But it seems like 
some people would like to choose Magma, if Magma can be a viable option 
for them. Chris may like Avi's 80% solution above. I don't know.

I am surveying my options. So this interests me.

Jimmie Houchin



More information about the Squeak-dev mailing list