Tool environments (was Re: [squeak-dev] Re: Hudson Status?)

Frank Shearar frank.shearar at angband.za.org
Wed May 4 07:59:29 UTC 2011


On 2011/05/04 03:59, Igor Stasenko wrote:
> On 4 May 2011 04:30, Chris Cunnington<smalltalktelevision at gmail.com>  wrote:
>> Hi Casey,
>>
>> Your offer to help with a Hudson server is valuable. And it will likely be
>> very valuable in the near future. I'm responding, because I don't want your
>> potential contribution to be ignored.
>>
>> There are some arguments that need to be worked out "in camera" before the
>> way forward is clear. You're an excellent programmer. Let us hammer out some
>> of the details first. The topic of modularity has come seemingly all at
>> once, but this plan has been waiting a year and this convocation of the
>> Squeak board has a year to get things set on the right course.
>>
>> For what it's worth, the video you're looking for is here:
>> http://www.youtube.com/watch?v=n4I7fSVNX2A
>>
>> If you go back to September Squeak postings you'll find I was one of it's
>> earliest proponents. And although I can't find the post right now, Stef
>> rejected this as a model for namespaces in Pharo.
>
> I can't say for Stef, of course. But i know that Pharo are slowly (but
> steady) drifting towards improved
> and modular model.
> Which means , better tools support for improved modularity. Package
> management support ,
> and eventually namespaces.
>
> Currently the work in progress going on, with main idea to abandon a
> concept of having single namespace in system (which is a system
> dictionary). It requires a bit ;) of  effort - making sure all tools
> working with 'environment' object, and not assuming that there are
> single global environment which they can access directly.

I'm not going to address modularity much, and stick to a much smaller 
issue, and one that affects me rather more directly: this environment 
object is a _great_ idea, and I've been meaning these past few days to 
look into what we have and what (if anything) I might need to do to it 
to get what I want... which is the ability to test Browser, Monticello 
stuff, etc., that affects the system.

Chris Muller & I both recently raised the issue where testing "system 
affecting software" - things that add/remove classes for instance - is 
sub-optimal: you end up generating loads of noise in the system (like 
the Unnamed changeset), and if tests fail you damage the image (by 
leaving half-instantiated or half-cleaned or half-manipulated junk).

Let's just not revisit the flame wars of old (to which Igor refers) and 
forget to do real, actual things that can benefit us regardless of what 
kind of namespace (if any) we want.

Increasing the testability of core browsers, CI servers are valuable 
things in and of themselves.

frank

> Only after this is done, it will be quite easy to introduce new
> namespace model.
>
> (concerning that video, i like the idea of having namespaces.. but
> browser with 5 list panels is too much)
>
> And concerning namespaces in overall, if you don't know it already, it
> is really hard to convince all smalltalkers, which
> namespaces model is better to employ .. its been a discussion about
> them over a years.. and none of them were actually
> lead to solution which is then implemented and forced into our images :)
>
>>
>> I don't want to say "We'll get back to you" and sound like a jerk, but
>> there's a screaming match brewing that the SOB that needs to resolve in
>> private before we can say what the next step is.
>>
>> Please bear with us,
>>
>> Chris
>>
>>
>
>
>




More information about the Squeak-dev mailing list