[squeak-dev] Re: package universes, sake/packages, (first time) user experiences, etc.

Keith Hodges keith_hodges at yahoo.co.uk
Fri Dec 5 21:42:31 UTC 2008


Edgar J. De Cleene wrote:
>
> On 05/12/2008, at 16:32, Greg A. Woods; Planix, Inc. wrote:
>
>> Squeak definitely needs a good strong leader, or at least a cohesive
>> leadership, with a good and hopefully popular vision of where the
>> core is going and how it's going to get there, and I think now with
>> the "basic" default image being one without everything pre-loaded
>> this vision has to stretch out over the basic package management
>> issues too.
>>
>
> But we don't have a good strong leader.
No we dont need a strong leader. We tried that one before and what we
got was an uncommunicative invisible ne disappearing leader.

What we need is people who are willing to contribute positively to the
positive forward thinking vision that we already have got. There is a
vision for 3.11, and there is a vision for 4, and 5. For 3 and 4 we have
a fairly bold and strong philosophical line, of making as many things
work for as many people as possible. And providing the tools required in
order to move forward.

Previous release teams identified the need for tools, for atomic loading
for example, then focussed primarily on the image once more, plugging
away with the old tools, and complaining all the way about the tools.

While the pharo faction is hammering away on the core image, its roughly
the same team who years ago complained that their work was hard because
the tools (i.e. Monticello) werent good enough for the job. They also
announce... "we are going for a more modular image"... and they have no
coherent strategy or effective tools for explaining how once I have a
more modular image I can (un)load things back in again and be sure that
they will work.

For years and years and years the difficulty of doing things particularly.

1) loading stuff getting dependencies right so things work
2) writing stuff that works and is loadable in many images
3) Providing feedback to the community about what works where and what
doesnt.
4) unloading (forget it)

These have been the biggest NEEDS in the community for those of us who
are using squeak for real work. They have ALL been addressed in the
ongoing 3.11/3.x tools effort.

Another thing that has been difficult has been migrating large codebases
from one image to another because the tools were different in each
squeak version. At one time I was working on projects in 3 different
squeak versions simultaneously, while maintaining 6 further images
because of the bugs in monticello1.

I dont need a better kernel I would like one, but I dont need it (YET).
Its a complete red herring until I can build production images that work
in less than 2 weeks. (thats what it took me using universes) Without
having to email every package owner, and twist their arm to fix things
or add my fixes to their code because I am using a more recent image
than they are.

Our goals are to make things better for everyone. The pharo team have no
such goals to ensure that their wonderful improvements to the image are
as widely useful to the community as possible. Any tools they produce
are in the first instance for their image only.

Did you know that someone in the Pharo camp recently merged FileList and
FileList2. Is it just me or could that improvement be useful to
everyone, whether in etoys or croquet. There is nothing in the Pharo
mindset or toolset that enables this to happen by default.

In contrast, for us, that is our number one goal. With the work that we
have been doing for 3.11, all the tools work for all the people in
theory. With the help of Installer, Sake/Packages and LevelPlayingField,
that FileList-improved could be made available to everyone, croquet,
etoys, sophie, not only that it can be loaded atomically too. We have
and are building the TOOLS to achieve this now.

Does that example help you see the difference of what we are trying to
achieve. We can and will catch up with these "exclusive visionaries", we
can pick the best of SqueakLight/FunSqueak/Pharo, but we can do it in a
coherent, considered "for as many as possible" manner.

The 3.x era is drawing to a close, what we need is more coherence, and
stability, and better tools.

The 3.11 team has a philosophy of improving the tools for everyone. And
if you look carefully a lot has happened on the tools front in the past
18 months.

There is a coherent strategy in place, and a vision, even if that vision
is a long time in coming, it is coming

volunteers are welcome to join us on squeak irc

Keith



More information about the Squeak-dev mailing list