[ANN] Magma 1.0gamma, on SqueakMap

chris at funkyobjects.org chris at funkyobjects.org
Thu Feb 6 06:28:06 UTC 2003

Ah, it feels good to come out of my cave to finally offer what appears
to be a usable version of Magma!  Multiuser is working*,
MagmaCollections are working*, and there are enough utilities and admin
functions for me (and you!) to write real programs.

For those that are interested in performance benchmarks, I put together
a commit test.  It was actually kind of fun to set up and run, and with
quite a bit of anticipation on the results.  Basically, for LAN, you get
about 15 commits per second on my Intel 700MHz laptop client.  For
local, 100 commits per second.  To read about it,

I tell you.. Six months ago, I was feeling a little burnt out on this
project.  I just couldn't wait to get done so I could do something else.
 It's been a long, trying journey, but patience and persistence have now
been rewarded with.. well more than just persistence!  :-)  Now, not
only do I get to be "done" I also get to start living my dream where
Smalltalk and large-models are at my disposal, and potentially even in
the company of other users!  I'm very excited.

* (did you notice the dreaded asterisk!)
So what condition is Magma in now?  Exactly how stable is it?  Well, the
SUnit tests are strict, especially on the core code.  Core functionality
seems to be very solid "gamma" quality at worst.  I'm not calling it
"Magma 1.0gamma" just because its fun to say.

Over and above the SUnit tests, I had a beta period where there was a
high-volume "task" it eventually worked through.  I'll elaborate on that
later.  And despite their complexity, MagmaCollections seem to be rock
solid due to the outrageous SUnit's they now endure, without flinching.

Overall, my feeling is that this may actually be usable.  Having said
that, keep in mind that I've only tested in Windows so far, and I have
yet to write my first "real" program using it.  But my suspicion is that
my "real" program isn't going to come close to the stress of the
SUnit's, and they all run multi-user.

Beware Windows 98!  I have discovered that Windows 98se can't seem to
handle being a server with consistent performance.  They have some kind
of problem in their TCP stack where the listening socket tends to
degrade, and it holds up requests getting to the server, slowing it
down.  It is not a Magma problem, nor is it a problem on Windows 2000,
which seems ready to serve rock-solid.

Enjoy it!

 - Chris

More information about the Squeak-dev mailing list