[squeak-dev] I went to Biloxi and all I got was a Tshirt
Chris Cunnington
smalltalktelevision at gmail.com
Mon Mar 19 15:03:57 UTC 2012
http://twitter.com/#!/metacello/statuses/173496047207849985
http://ss3.gemstone.com/ss/FileTree.html
https://github.com/finworks/smallsource
Here in Biloxi at STIC, formerly Smalltalk Solutions. It's nice to go
from Toronto at 11 degrees to Biloxi at 25 degrees. This is not Edinburgh.
The casino has a weird policy on wifi so there is none in my room. There
is an Ethernet jack and a cable bolted to the desk. It is no longer
bolted. The details of why it is no longer bolted or how that nylon
twist tie liberated itself I can neither confirm nor deny.
Met most of the people I'll see for the next few days last night. So the
introductions are done, I'd say. I'm going to a presentation in an hour
on Dart by Eric Clayberg, who works on Dart at Google. Should be
interesting in an academic way, as nobody will ever use Dart for
anything. Tomorrow is a talk by Sam Adams at IBM called Big POOP. They
are having fun with that name - Parallel OOP.
I thought Smalltalk Directions, the parallel track, the academic track
was going to be a cavalcade of stars. It has exactly one panel. I
thought it was going to be a cavalcade, as there are about fifteen
famous names on the paper vetting committee, none of which are here. If
I understand correctly the Smalltalk Academic Advisory Council will have
its one panel tomorrow with Andrew Black, Ralph Johnson, David West, and
Roger Whitney. That is not to be missed. I guess the panel is about the
future, so it could be about anything. Several people on that panel are
destination speakers, I'd say.
Chris Muller and myself had an interesting conversation with Dale
Henrichs last night. Dale's doing a talk called Practical Git For
Smalltalk. He want's to disintermediate squeaksource and squeaksource3
and use GitHub for Smalltalk. Instead of saving to an mcz file, you'd
save to a directory structure with a chunk file for each method. The
snapshot in an mcz file would no longer be a binary, a serialized object
graph, but .st files in directories. If I understood correctly, the
metadata glue would be JSON.
Seems he had a light bulb moment reading about Otto Behrens at
Finworks.biz, who has a process of doing just that. Otto's git page is
above along with the SS3 location and the Metacello repository
conversation. I tried loading ConfigurationOfFileTree in Pharo 1.3 and
it failed. Then I tried to load ConfigurationOfFileTree in Pharo 1.4 and
got the same message: 'The symbolic version #stable is not defined in
ConfigurationOfFileTree for the current platform'. OK. I guess I'll ask
Dale about it.
The other idea Dale was mentioning about Metacello was to remove its
dependency on Gofer or Installer, so that it could be directly
scriptable.He says instead of a whole inhalation of classes when
Metacello is used, there would be only two living permanently in the image.
Seems I learned why MetacelloBrowser was a non-starter. For the sake of
speed he used the class browser and then OmniBrowser to get Metacello
running. But what Metacello needs he says is a wizard, where the user
makes a few choices over a few pages. I can't say I'm familiar with any
such thing in Smalltalk.
OK, looks like if you resist the temptation to use Metacello and use
Monticello to load MonticelloFileTree and MonticelloFileTree-Core from
SS3, then they load in Pharo 1.4. And there are class comments. I can
read those. And talk to Dale. Look at that. They load into Squeak4.3
without exploding. That's good.
Well, you know what I know for the moment. I'm off to see a presentation
on a brilliant language that will only ever work on Chrome. More coffee!
Chris
More information about the Squeak-dev
mailing list
|