Hi Göran -

I'm a big fan of builds in production systems but I'm a little at a loss 
how exactly you are using Installer and how it helps you address 
particular needs. At Qwaq, we use Monticello exclusively and the way 
we're working is that we have our own versions of *all* the Monticello 
packages we depend on that we modify when we have to. What it means is 
that we have a consistent development process, don't rely on any 
external servers and that (for example) integrating a fix is as simple 
as loading it and committing to the repository which is consistent with 
the rest of the development activities. We can do this both in our own 
code as well as in the base code and we don't need to wait for anyone 
before the next product build can go out.

In Gjallar, how do you deal with the same set of problems (making a 
modification for the product, integrating some external modification) 
using Installer? In particular how do you deal with issues like having 
to rely on external servers (which might be on the other end of the 
world, down and out etc), the ability to make some changes quickly for 
the product to ship etc.

I have never found a production setup which did not require to 
ultimately have your own copies of the code on your own servers in which 
case you're quickly getting to the point that a single homogeneous 
environment is advantageous (like using MC+configs, or using an update 
stream) and where a mix of these ultimately only gets you into trouble. 
The reason why I like DeltaStreams is that it sounds like an approach 
that allows one to reconcile the (necessary) local mods with other 
changes more easily than could be done by MC or updates. But I don't 
think Installer is quite that and so I'm not sure how it helps in the 
process. An experience report would be most welcome.

The other question here is that I don't understand how Installer has 
helped you moving between versions. Having moved Croquet from 3.2 to 3.4 
to 3.6 to 3.8 I have found each version a completely unique challenge 
and I cannot imagine that any particular tool would have helped me 
getting this work done more efficiently than the (extremely painful) 
process of loading, having it break, figure out where exactly that 
interface went, rinse and repeat. Anything that could help with this 
process would be hugely beneficial.

Cheers,
   - Andreas

Göran Krampe wrote:
> Hi all!
> 
> First of all - great thread!! :) I could ask questions and comment on 
> hundreds of things but I am letting the current people hash it out for 
> now. One thing I would like someone to do though is to "distill" this 
> thread into at least a Swiki page on the Squeak Swiki! :)
> 
> But a bit of comments anwyay:
> 
> Keith Hodges wrote:
>> Jerome Peace wrote:
>>> I personally do not want to have to understand what is on the pbwiki 
>>> or to navigate keith's new ways of doing things in order to play and 
>>> test out a new squeak image.
>>>
>>> What unsettles me at the moment is that two very powerful programmers 
>>> are taking 3.11 in some very new directions relative to what the 
>>> community is used to.
>>>   
>> The alternative was officially nothing. I piped up when the board were
>> considering cancelling 3.11
>>
>> So anything is better than nothing, perhaps?
> 
> Indeed and I have repeatedly asked for a clarification on the 3.11 
> status and how "official" it is! Now I hope it is FULLY official and 
> that Matthew (?) is the release team leader.
> 
>>> While I have a great respect in Matthew's judgment and ability to 
>>> explain what he is doing, I have found from experience that Keith's 
>>> notions are more of a gamble.
>>>   
>> So far, I am quite pleased to say that everything I have put my hand to
>> has worked really well.
>>
>> But this comment indicates to me that you really dont "get it" yet.
>>
>> The whole deal with this 3.11 project is not about delivering an image,
>> its about addressing a need, through putting a philosophy into practice
>> across the board. The 3.11 goal is to showcase the tools that make that
>> philosophy possible. While the tools are not ready there is no 3.11
>> (fortunately the tools are getting there and there will be a 3.11)
> 
> Many of us who have been around for a while know that we have tried 
> numerous approaches over the years and we know that many of these have 
> failed due to various reasons. For example, the idea of "harvesting" and 
> having so called "harvesters" failed because it only led to a few people 
> burning out (in the very early years it "worked" because we had paid 
> people doing the harvesting).
> 
> So such a model is not viable IMHO. We need to move and try other routes 
> - and I for one applaud the current effort in 3.11 and will try to chip 
> in wherever I can.
> 
>> The need is definitely there, and the philosophy aiming to meet that
>> need has been operating well for almost a year now. That's not a gamble
>> at all, its already happening.
>>
>> Edgar has delivered image after image, but does that help anyone in the
>> long run really. It  doesn't help me. I have production code and I don't
>> have time to spend a month moving it form one image to another manually,
>> without any tools to help, broken MC, broken Universes etc. It doesnt
>> help us move forward in the future to something like Morphic 3.0, or COG
>> for which atomic loading is likely to be essential.
>>
>> Real World Example:
>>
>> As an example, Gjallar was working in 3.8, there is no technical
>> compelling reason to move the huge code base over to 3.10. It doesn't
>> offer any must have new features. The only reason for moving is to be
>> able to keep up for the sake of it. So into this situation comes
>> Installer, Gjallar migrates to use Installer for its build scripts
>> (July2007). Once Installer is used, the build script can be run in 3.7,
>> 3.8, 3.9, 3.10, Sophie, Croquet or Etoys. Installer proves the common
>> ground that is essential to move forward, even though that move didn't
>> take place for almost a year, the Gjallar team knew that they were no
>> longer a fork, because they had the tools to make keeping up possible
>> and straight forward.
>>
>> So did Gjallar move to 3.10 because of the 3.10 image features, or
>> because Installer and LevelPlayingField helped make it a smooth
>> transition?  Gjallar is a fork no more. Which of 3.10/Installer is
>> really contributing most to moving the community forward? I would say
>> that Installer is doing a damn fine job for a little tool.
> 
> Yes, we made the move just recently of Gjallar over to 3.10 and adjusted 
> our Installer script to use LPF etc. It worked really well and we did it 
> mainly because we don't want to be left out on the improvements/fixes 
> pouring in. All in all Installer is a great tool making "image building" 
> quite easy. Combined with Sake/Packages (modulo having not used it yet) 
> I presume it gets even more powerful.
> 
> A trivial example from Gjallar: We include fixes available on Mantix 
> using Installer "oneliners". We don't need to wait for someone else to 
> harvest it, get it into the image etc etc.
> 
> [SNIP]
>> The important thing is that all of the contributions to that 3.11 are
>> also available to all other image users. So you don't have to wait for a
>> 3.11 image to partake of the new wine.
>>
>> Its not really a gamble its a coherent strategy to implement what Goran
>> had as a vision, multiple update streams, but in a different form.
>> Different experts can contribute different tasks managed in Monticello.
> 
> Yes, we share the same understanding of where we are right now - the 
> Squeak world is already "forked" in several directions. We need to get 
> mechanisms in place to cope with that. I still hope to be able to move 
> DeltaStreams forward (time, time, time...) but the important thing is 
> that we are several who share the understanding of the core problem.
> 
>> Dont Panic
> 
> Hehe, I like it! :) Remember folks - we are all in this for fun! Matthew 
> and Keith (and several others of course) have made huge contributions 
> over the last year, and I am extremely grateful for that.
> 
> I hope to be able to pull my share mainly related to SM.
> 
> regards, Göran
> 
> 
>