>>
>> I think we should consider cribbing from the model that Linux distros
>> use:
>>
>> Load (more precisely: use -- we might not be able to load it) a /
>> specific/ version of the Squeak tool, and then apply a separate
>> (Pharo-
>> maintained) set of patches to make it Pharo compliant.
>
> but only the developer of the tool can do that.

No, quite the opposite:

The Pharo maintainers maintain the Pharo compliance patches for every  
tool we import, the tool developer should not be bothered with these  
-- should not even see them.

Only a subset of these patches -- the bugfixes created in the Pharo  
community -- should be pushed up to the tool developer.



This way both worlds could get the best of both worlds:

The tool developers still get bugfixes posted despite their original  
implementation possibly being/becoming incompatible with Pharo.

The Pharo community gets the freedom to become incompatible with  
Squeak without loosing the benefits of ongoing tool development on  
Squeak.


I am not saying this is the only solution, just describing what is a  
'field proven' model in the Linux community which seems to perfectly  
implement the stated goals of Pharo.



>
>> [...for version management]
>>> The vision is bytecode loader
>>
>> Pleas don't go there. Byte-code loading is a very nice-to-have in
>> deployment scenarios (lazy plugin loading, live update delivery to
>> deployed apps etc) but if there is one thing that using VW's Store  
>> has
>> taught me it is that it definitely should not be part of a code
>> repository that is used for development code version management.
>
> This is in VW store is bad. I do not see why a package MCZ could not
> ship with
> a byte-code synchronized version of the source. Then we could have a
> stripper
> or a synchronizer that recompile the source code and produce new
> bytecode.


In my experience it 'poisons' the repository with packages that load  
fine as bytecode but fail to load as source.
Mostly in 'auto brain surgery' scenarios where the /source/ should  
contain more load ordering hints (but never attains that level because  
the binary version hides these bugs).

I reiterate that VW shows that source loading is fast enough -- even  
for rather large updates.

So in my opinion 'fast enough' source loading should be attained  
first, and bin-loading should be added later as an option (and  
probably not in the code repository but perhaps in a cache in front of  
such a repository, or maybe sideways as part of a build process).

>
> we will start by fixing packageInfo

:-) right!



R
-


_______________________________________________
Pharo-project mailing list
[email protected]
http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project

Reply via email to