I do not know.
What would like is
        - well documented packages classes with tests and that we can load what 
we need.
        - small core
Stef

On Jan 17, 2011, at 7:03 PM, Hernán Morales Durand wrote:

> Hi, and thanks for the clarification Levente,
> 
> This points me to a question, since it seems KeyedSet was removed
> because at that time no users were found. Is that the criteria (i.e.
> no users) for removing a class from the Core image?
> 
> Assuming you have NO resources for checking use of classes by the
> users, and core developers have NO care/time/energy for annotating
> their changes (or there is no tool to do it or nobody uses it), I have
> this idea:
> 
> You continue deleting classes as if nothing happens, BUT let users to
> vote for a request an addition or NOT removal. Then you win on two
> aspects:
> 
> -First you get a free log of what has been removed
> -When votes reach a consensus value, that you may seriously consider
> including whatever is in question.
> 
> How about tagging issues like [RFI] (Request for Inclusion) ? And
> every comment counts like a vote.
> 
> Opinions?
> 
> Hernán
> 
> 2011/1/15 Levente Uzonyi <[email protected]>:
>> On Sat, 15 Jan 2011, Hernán Morales Durand wrote:
>> 
>>> Hi Stef
>>> 
>>> 2011/1/15 Stéphane Ducasse <[email protected]>:
>>>> 
>>>> On Jan 14, 2011, at 6:24 PM, Hernán Morales Durand wrote:
>>>> 
>>>>> Try PluggableSet
>> 
>> Note that KeyedSet has nothing to do with PluggableSet. PluggableSet is a
>> set where you can specify the hash and equality of two objects with blocks.
>> This gives you great flexibility without creating a subclass of Set.
>> KeyedSet is like a dictionary, but the keys of the objects are not stored in
>> the data structure, because the objects know it themselves. This means that
>> the data structure will have a smaller memory footprint.
>> 
>> 
>> Levente
>> 
>>>>> 
>>>>> 
>>>>> http://www.mail-archive.com/[email protected]/msg22890.html
>>>>> 
>>>>> Is there a way for you guys to deprecate deleted classes or mantain a
>>>>> record of deleted behaviors? (IIRC VW did this from 2.x versions)
>>>> 
>>>> We try but we do not have the engineers of VW.
>>>> Now find somebody that pay us just 3 engineers full time for two years
>>>> and you will not
>>>> recognized pharo and its process.
>>> 
>>> I know we are very few, but what about this idea, instead of just
>>> deleting a class we may adopt these simple rules:
>>> 
>>> 1 - If deleting a class add the deleted class name somewhere (in a
>>> method?)
>>> 2 - If renaming a class, make the class to know its previous names
>>> like #previousNames
>>> (Norberto Manzanos implemented this for his own package management and
>>> I think is a great idea)
>>> 
>>> doesn't require tools, just some consensus and maybe a modified
>>> #deleteClass to enforce this "policy"
>>> 
>>> it sounds feasible for you? what do you think?
>>> 
>>> Hernán
>>> 
>> 
> 


Reply via email to