[Squeakfoundation]KCP & 3.6

Doug Way dway at riskmetrics.com
Wed Jun 18 20:04:10 CEST 2003


goran.krampe at bluefish.se wrote:

>Hi Stephane, Andreas and all!
>
>Instead of quoting a lot of text etc I try to keep it to the point:
>
>- Keep on cleaning Stephane! Sure, sometimes people will disagree with
>selected proposed actions - but we are *all* grateful for the work being
>done, including Andreas I am sure. So please, don't be discouraged. The
>progress so far is great.
>

Agreed!  (Belatedly chiming in...)

>- Regarding these "hard to place methods", placing #abandonSources in
>Image does sound perfectly reasonable to me... And I would also expect
>to find the #snapshot:andQuit: family of methods there (even though I
>don't really like the names of those...).
>

This seems to be the main subject of debate.  Whatever KCP decides to 
do, it would probably be good to at least post the proposed class names 
to the list before going through the trouble of implementing them & 
changing all the references.

(For my 2 cents:  I'd rather not have too many of these classes either.  
Also, I'm not crazy about using singleton instances unless there's a 
real possibility that there might be multiple instances.  Class-side 
methods seem more straightforward.)

>- Regarding the rest of these methods, there may be methods that really
>are hard to place - and in that case I agree that a Utilities class can
>be good to have, BUT:
>	- Placing them all in one class may need the use of class extensions so
>that it doesn't "hurt" the partitioning of the image. If we put a
>utility method for Morphic and a utility method for files in the same
>class called Utilities then we will create problems for minimal images
>without Morpic - unless we use class extensions.
>

True.  Although KCP isn't doing any image partitioning right now, so we 
don't have to worry about that just yet. :-)  But it's something to keep 
in mind.

As a general rule, I'm assuming that we are allowing the use of class 
extensions when partitioning the image.  (perhaps stating the obvious)

>	- Having utility methods in class SystemDictionary seems utterly wrong
>to me. Just listen to the name - SystemDictionary. What on earth tells
>the novice that this is a place for Utility methods? Nothing IMHO.
>

I think everyone agrees with this... this was part of Nathanial's 
excellent summary.

>It is
>also confusing for the novice that the methods are in "SystemDictionary"
>but you run them by sending messages to "Smalltalk"... But the class
>name Utilities or even UtilityMethods with the methods on the class side
>is much better I think. If we indeed are talking about functions then I
>think it is appropriate for them to be on the class side.
>  
>

I tend to agree here too.

On the subject of KCP, if we want to get another batch of KCP updates in 
before 3.6 goes beta, we should probably do that soon.  I'll leave it to 
Daniel to coordinate harvesting/approving that.  (Realistically I think 
we need to have one harvester in charge of these types of cleanup 
projects... it would be difficult to rely on the regular harvesting 
process to individually approve a bunch of cumulatively dependent 
changesets.)

Speaking of 3.6beta, technically the planned date for this is in two 
days, this Friday the 20th.  But I think we could be a little flexible 
with that date, since we're now finally moving with harvesting stuff... 
definitely move to beta sometime next week though.  We should still keep 
our final date of August 1st, though, unless we agree now to postpone it 
for some reason.

- Doug Way




More information about the Squeakfoundation mailing list