[Module] Ginsu for 3.1 (was: extremely dissapointed)

ducasse stephane ducasse at iam.unibe.ch
Sat Sep 22 19:37:08 UTC 2001


hi Daniel

have you any idea where I could find LinkedMessageSet this is Ginsu but not
in sq31 neither in your rb version.

> I want you to continue. If the ModSqueak work doesn't get in because it
> hinders our development style, that's ok. If it doesn't get in because
> nobody bothered to make it work in a current Squeak, that's a pity. Yell
> if you need help with anything.

I'm continuing. I was just saying that not replying to email and key
questions is frustrating. Because as a scientific guy I would like to
understand 
    - if namespace do/or not (I'm favor or not) have to be coupled with
module.
    - do we need a scoping mechanism or not
    - how teamV/Envy style hampers open distribution of fix and enhancement?
    ......

> 
> (I get this funny feeling that maybe modules are a cursed topic. I never
> really liked what they did to other languages, though they're better
> than one huge source file. And here it seems everyone working/talking on
> modules starts behaving strangely. This began way back with the HUGE
> disscussions, and upto the current (not fun) state of affairs).

There are not. There are the only way to manage complexity and decentralized
Squeak and make even Squeak multiples: we can think that pretty soon I will
be able to download (build from a description) an image of the latest SqC
squeak or the latest SqueakScript, or .....
> 
> Anyway, I'm not sure I know what each part of the SWT work consists of
> exactly. Assuming you port Ginsu to 3.1, this will include -
> * The ModSqueak module loader
> * Hierarchies of modules, clusters so forth to hold the loaded code,
> whether installed or not.
> * Ginsu itself, a tool for refactoring this hierarchy.
> Is this right?

>From what I see Ginsu is:

- a model for managing code (I would like to get rifd of it to keep the
causal connection of the smalltalk meta model may be we should just
introduce GlobalVariableDefinition and use it)
getting rid of it would let normal tools work and we would have just to
adapt the system to use the globalVariableDefinition and the other that are
missing, like ClassDefinition (because waht is displayed in a browser is
dynamically reconstructured fomr the object representing the class).

- 4 classes representing the module

- a module builder that classify according to a spec the contents of the
image into modules and submodules

- a set of error related to the repository, i.e., loading and installing
modules

- a set of tools to manage that


> Do you need beta testing?
Certainly but not now, I'm fighting with it to get it loaded and try to see
if I can fix the tests that do not run.


> I want to continue evolving the RB in something that's more module
> aware.

So wait that the module concept exist. If you wantto have fun look at the
classification idea developed by roel www.iam.unibe.ch/~wuyts/
because this is a really interesting way of been able to manipulate ***any**
kind of software artficats such as inspected object, categories, modules,
protocols, methods, method senders.....

> An experiment with loading modules from SCAN sounds fun (as soon as I
> can get a module system with a usable loader, and tools for refactoring
> the modules).

Sure and joseph and hans discussed at ESUG and their work is really
complementary 
> Thanks,
> Daniel
> PS - 
> ducasse stephane <ducasse at iam.unibe.ch> wrote:
>> Hi
>> 
>> I just want to tell that the atmosphere that is on the emails related to
>> modules is really disappointing. I'm porting Ginsu to 3.1 but I'm really
>> wondering if I should not spend my time on other urgent things that I have
>> to do for my future.
>> 
>> For me I do not care if the work of Joseph or Henrik is one chosen. For me
>> only the quality is the issue. I think that the topic is so important for
>> the future of Squeak (that will stay a nice blob if we cannot ensure a
>> component architecture that supports the distribution and open source
>> distribution).
>> 
>> So may be joseph with his natural arrogance pissed of a lot of people. Ok I
>> can understand that.
>> 
>> I would appreciate that people react to the questions I ask about module and
>> the implementation of Henrik and I'm really sad. I have the impression that
>> the head of Henrik is inflating so take care....
>> 
>> I think that this is easy to say that a design does not support open
>> distributed whatever and this is easy to reinvent the wheel. So good luck.
>> 
>> Stef
>> 
>> Sorry I'm really pissed of by some of the emails I read. So may be I'm
>> misinterpreting but at least you know my feelings: discouraging!!!
> 
> 





More information about the Squeak-dev mailing list