hi guys
We think that this is important that we build some plans for 3.8. As we already mentioned it, if someone needs urgently 3.8, he should do it. We (marcus doug and me) would like to rest a bit and get some real stuff done.
We discuss with diego and others about a good date and the idea would be to have 3.8beta the 15th of november and 3.8 ready for 15 of december. (no gamma phase this time). We lose too much time with that. Just some tentative images.
In addition to m17n, what we would like to see with this version: - services (so please have a look and comment) - shout (already ready to go as package) - OmniBrowser (with refactorings :)) - refactorings integrated with services in all the browers - a lot of fixes (so please continue to review and give your point of view!) - a lot of tests we need more!!! Really everybody can write tests. - excellent diego look - more AST and compiler cleans (I should check with marcus): the AST of RB + visitor is already in so consider to see if you can use it for your tool. - new preference browser - clean of the system dictionary (see my next email) - SM20 with configuration (yes yes) - we are thinking having SmallLand as full image but we will see if diego has still the steam for that. But indeed now the unstable stream idea is really helping us. I certainly forgot stuff. But this is important that we have a clear date in mind, so as you may have noticed the harvesting is going much faster than ever and ned is not back yet :)
Again we would need a Etoy and multimedia guide. I repeat We need a Etoy guide and Multimedia guide. I suggest the cool ones in the community to try SqueakUnstable (which is quite stable since now).
Stef
By the way Traits are ready to get in but I forgot to mention them, strange. Alan always said to nathanael that he would like to have a smalltalk with traits, so we have a good opportunity. Now this would be important to know if someone is against.
Stef
On 28 sept. 04, at 16:34, stéphane ducasse wrote:
hi guys
We think that this is important that we build some plans for 3.8. As we already mentioned it, if someone needs urgently 3.8, he should do it. We (marcus doug and me) would like to rest a bit and get some real stuff done.
We discuss with diego and others about a good date and the idea would be to have 3.8beta the 15th of november and 3.8 ready for 15 of december. (no gamma phase this time). We lose too much time with that. Just some tentative images.
In addition to m17n, what we would like to see with this version:
- services (so please have a look and comment)
- shout (already ready to go as package)
- OmniBrowser (with refactorings :))
- refactorings integrated with services in all the browers
- a lot of fixes (so please continue to review and give your point of
view!)
- a lot of tests we need more!!! Really everybody can write tests.
- excellent diego look
- more AST and compiler cleans (I should check with marcus): the AST
of RB + visitor is already in so consider to see if you can use it for your tool.
- new preference browser
- clean of the system dictionary (see my next email)
- SM20 with configuration (yes yes)
- we are thinking having SmallLand as full image but we will see if
diego has still the steam for that. But indeed now the unstable stream idea is really helping us.
I certainly forgot stuff. But this is important that we have a clear date in mind, so as you may have noticed the harvesting is going much faster than ever and ned is not back yet :)
Again we would need a Etoy and multimedia guide. I repeat We need a Etoy guide and Multimedia guide. I suggest the cool ones in the community to try SqueakUnstable (which is quite stable since now).
Stef
Hi Stef,
By the way
Traits are ready to get in but I forgot to mention them, strange. Alan always said to nathanael that he would like to have a smalltalk with traits, so we have a good opportunity. Now this would be important to know if someone is against.
Stef
I'm not against traits at all but I have comments to do (note: I sent this comments to people behind traits months ago and I got no answer)
What I don't like in traits is the procedure to resolve conflicts. The politic to create a flat view of all the traits + super class produces some of the problems we find in multi-hierarchy inheritance.
Concretely I would like to hear from you (the traits team) why a flat-everything model is better than (my favorite) the mixins in Strongtalk [1]. As long as I understand, this model shared the benefits of traits avoiding the problems when conflicts occurs.
BTW, What about a first version of traits/mixins/callAsYouWant where conflicts are not allowed? In this case we can start to move from the single-hierarchy while we find a way to resolve the conflicts.
Cheers,
-- Diego
[1] http://www.cs.ucsb.edu/projects/strongtalk/big/mixins-paper.ps
hi diego
I will put online a paper that we just have finish to write that explain why mixins are a problem (even gilad told us that we were right!).
I will send you the pdf.
I'm not against traits at all but I have comments to do (note: I sent this comments to people behind traits months ago and I got no answer)
What I don't like in traits is the procedure to resolve conflicts. The politic to create a flat view of all the traits + super class produces some of the problems we find in multi-hierarchy inheritance.
The idea is with traits the composite entity is responsible to compose and resolve the conflicts. Read the paper I will send you and let us know what you do not like.
Concretely I would like to hear from you (the traits team) why a flat-everything model is better than (my favorite) the mixins in Strongtalk [1]. As long as I understand, this model shared the benefits of traits avoiding the problems when conflicts occurs.
No. Not at all. Explicit conflict is much better than implicit one because you have the control while wiht implicit control you have to turn around. You will understand when reading the paper.
BTW, What about a first version of traits/mixins/callAsYouWant where conflicts are not allowed?
This makes no sense.
In this case we can start to move from the single-hierarchy while we find a way to resolve the conflicts.
Cheers,
-- Diego
[1] http://www.cs.ucsb.edu/projects/strongtalk/big/mixins-paper.ps
After fighting with the system I put the paper available for everybody: http://www.iam.unibe.ch/~scg/Archive/Papers/ Duca04wtoplastraitnotfinal.pdf
stef
On 28 sept. 04, at 17:26, Diego Gomez Deck wrote:
Hi Stef,
By the way
Traits are ready to get in but I forgot to mention them, strange. Alan always said to nathanael that he would like to have a smalltalk with traits, so we have a good opportunity. Now this would be important to know if someone is against.
Stef
I'm not against traits at all but I have comments to do (note: I sent this comments to people behind traits months ago and I got no answer)
What I don't like in traits is the procedure to resolve conflicts. The politic to create a flat view of all the traits + super class produces some of the problems we find in multi-hierarchy inheritance.
Concretely I would like to hear from you (the traits team) why a flat-everything model is better than (my favorite) the mixins in Strongtalk [1]. As long as I understand, this model shared the benefits of traits avoiding the problems when conflicts occurs.
BTW, What about a first version of traits/mixins/callAsYouWant where conflicts are not allowed? In this case we can start to move from the single-hierarchy while we find a way to resolve the conflicts.
Cheers,
-- Diego
[1] http://www.cs.ucsb.edu/projects/strongtalk/big/mixins-paper.ps
Hi stéphane and fellow squeakers,
We think that this is important that we build some plans for 3.8. As we already mentioned it, if someone needs urgently 3.8, he should do it. We (marcus doug and me) would like to rest a bit and get some real stuff done.
You guys deserve it! :-)
We discuss with diego and others about a good date and the idea would be to have 3.8beta the 15th of november and 3.8 ready for 15 of december. (no gamma phase this time). We lose too much time with that. Just some tentative images.
We would like to suggest that we close 3.8 asap and only allow fixes. There are too many things that are broken through recent stuff, especially m17n related. E.g., for me personally I still need to fully adapt Yaxo, Jabber and even some of the network code to m17n.
So our list:
In addition to m17n, what we would like to see with this version: - services (so please have a look and comment) - shout (already ready to go as package) - a lot of fixes (so please continue to review and give your point of view!) - excellent diego look - new preference browser - clean of the system dictionary (see my next email) - SM20 with configuration (yes yes)
Did we mention fixes? ;-) And tests are always welcome, not just for 3.8 ;-)
All other items are pretty deep system changes and IMHO would require a longer testing phase.
So the release date could be nov 25th (Thanksgiving) and oct 15th could be the beta stage (feature freeze) date.
So we (Bert and I) would like to suggest that until oct 15th enh etc relating to the above list can be posted into the unstable stream. After that date only fixes can go in. We would also like to suggest that we handle bug tracking and collection of fixes through mantis (http://bugs.impara.de).
Bert&Michael
On Sep 28, 2004, at 11:31 AM, Michael Rueger wrote:
So our list:
In addition to m17n, what we would like to see with this version: - services (so please have a look and comment) - shout (already ready to go as package) - a lot of fixes (so please continue to review and give your point of view!) - excellent diego look - new preference browser - clean of the system dictionary (see my next email) - SM20 with configuration (yes yes)
All other items are pretty deep system changes and IMHO would require a longer testing phase.
Just as a point of reference, OmniBrowser makes no changes to the base system at all, only additions.
If the plan is indeed to replace the existing system browser, I think OB should be included (yes Göran, as a package) in 3.8. Developers can choose to use the old browser if they run into problems with OB, and we'll benefit a lot from the wider testing OB will get from being in the release.
Colin
Hi mike
We discuss with diego and others about a good date and the idea would be to have 3.8beta the 15th of november and 3.8 ready for 15 of december. (no gamma phase this time). We lose too much time with that. Just some tentative images.
We would like to suggest that we close 3.8 asap and only allow fixes. There are too many things that are broken through recent stuff, especially m17n related. E.g., for me personally I still need to fully adapt Yaxo, Jabber and even some of the network code to m17n.
So our list:
In addition to m17n, what we would like to see with this version: - services (so please have a look and comment) - shout (already ready to go as package) - a lot of fixes (so please continue to review and give your point of view!) - excellent diego look - new preference browser - clean of the system dictionary (see my next email) - SM20 with configuration (yes yes)
Did we mention fixes? ;-) And tests are always welcome, not just for 3.8 ;-)
All other items are pretty deep system changes and IMHO would require a longer testing phase.
So the release date could be nov 25th (Thanksgiving) and oct 15th could be the beta stage (feature freeze) date.
So we (Bert and I) would like to suggest that until oct 15th enh etc relating to the above list can be posted into the unstable stream. After that date only fixes can go in. We would also like to suggest that we handle bug tracking and collection of fixes through mantis (http://bugs.impara.de).
Ok so we agree on that list. Because this is more than fixing m17n
Bert&Michael
On 28 sept. 04, at 19:07, Michael Rueger wrote:
stéphane ducasse wrote:
Ok so we agree on that list. Because this is more than fixing m17n
Almost. After reading your SystemDictionary proposal I would vote for postponing that as it is a *very* deep system change.
In fact not really because if we do it fast the end user or other programmer will not see it, while if we do it over a long period of time then we will have to change to point to the right place for a temporary moment. The point to have Smalltalk pointing to SmalltalkImage current and merging SmalltalkImage and SystemDictionary we would keep all the old but deprecated methods.
But you are right that we should postpone it, so may be for us just introducing namespace would be enough to let us work.
I could be persuaded to have OB included :-)
I do not really understand what you are implying: what do you want in the list because your email was really not clear?
We would like to suggest that we close 3.8 asap and only allow fixes. There are too many things that are broken through recent stuff, especially m17n related. E.g., for me personally I still need to fully adapt Yaxo, Jabber and even some of the network code to m17n.
So our list:
In addition to m17n, what we would like to see with this version: - services (so please have a look and comment) - shout (already ready to go as package) - a lot of fixes (so please continue to review and give your point of view!) - excellent diego look - new preference browser - clean of the system dictionary (see my next email) - SM20 with configuration (yes yes)
Did we mention fixes? ;-) And tests are always welcome, not just for 3.8 ;-)
All other items are pretty deep system changes and IMHO would require a longer testing phase.
mike can yoou clarify that list because you confused me.
We would like to suggest that we close 3.8 asap and only allow fixes. There are too many things that are broken through recent stuff, especially m17n related. E.g., for me personally I still need to fully adapt Yaxo, Jabber and even some of the network code to m17n.
So our list:
In addition to m17n, what we would like to see with this version: - services (so please have a look and comment) - shout (already ready to go as package) - a lot of fixes (so please continue to review and give your point of view!) - excellent diego look - new preference browser - clean of the system dictionary (see my next email) - SM20 with configuration (yes yes)
Did we mention fixes? ;-) And tests are always welcome, not just for 3.8 ;-)
All other items are pretty deep system changes and IMHO would require a longer testing phase.
So the release date could be nov 25th (Thanksgiving) and oct 15th could be the beta stage (feature freeze) date.
So we (Bert and I) would like to suggest that until oct 15th enh etc relating to the above list can be posted into the unstable stream. After that date only fixes can go in. We would also like to suggest that we handle bug tracking and collection of fixes through mantis (http://bugs.impara.de).
When I read that I understand that this list is ok for you (except SystemDictionary deep stuff). You removed OB and refactorings but I do not really see why since the AST+Parser are already in the image now and they do not impact the system.
By the way do you imply that you will do the harvesting of changes?
Stef
Hi guys!
A few short notes:
1. The list of stuff sounds fine by me. As long as stuff is added *as packages*, but I take that for granted. 2. SmallLand as full image? Hmmmm, doesn't sound like the same audience to me. 3. Traits - well, I need to hear a short summary of how it "affects" Squeak before giving it my definitive blessing. BUT don't get me wrong - I was REALLY impressed with the original OOPSLA presentation and I really like that we actually start moving Squeak into New Territory. 4. OB - sure, go for it. :) 5. SM2 with dependencies - yes, I agree. Working on it. In fact - it is the ONLY thing I have even remotely time to do now. :) A status report will arrive pretty soon on SqP.
Well, can't come up with more at this point. :)
regards, Göran
For traits go on SM and load it :).
Now when you edit a method in the debugger the changes goes in the right trait. When adrian is back from holiday he will certainly check the new version of OB. Basically traits do not change too much (even if we need to recompile all the class, we just need a new class above Behavior) and as the VM as some hard assumption on the instance variable we had to move them in PureBehavior. Beside that the changes are contained to Behavior, Class, ClassDescription and MetaClass.
Stef
On 29 sept. 04, at 21:18, goran.krampe@bluefish.se wrote:
Hi guys!
A few short notes:
- The list of stuff sounds fine by me. As long as stuff is added *as
packages*, but I take that for granted.
In fact I hope that we will be able to have a package for each category one of these days and this will be fixed.
- SmallLand as full image? Hmmmm, doesn't sound like the same audience
to me. 3. Traits - well, I need to hear a short summary of how it "affects" Squeak before giving it my definitive blessing. BUT don't get me wrong
I was REALLY impressed with the original OOPSLA presentation and I really like that we actually start moving Squeak into New Territory. 4. OB - sure, go for it. :) 5. SM2 with dependencies - yes, I agree. Working on it. In fact - it is the ONLY thing I have even remotely time to do now. :) A status report will arrive pretty soon on SqP.
Well, can't come up with more at this point. :)
regards, Göran
The version of traits on SqueakSource is more recent, we will publish a new one soon.
On 29 sept. 04, at 21:18, goran.krampe@bluefish.se wrote:
Hi guys!
A few short notes:
- The list of stuff sounds fine by me. As long as stuff is added *as
packages*, but I take that for granted. 2. SmallLand as full image? Hmmmm, doesn't sound like the same audience to me. 3. Traits - well, I need to hear a short summary of how it "affects" Squeak before giving it my definitive blessing. BUT don't get me wrong
I was REALLY impressed with the original OOPSLA presentation and I really like that we actually start moving Squeak into New Territory. 4. OB - sure, go for it. :) 5. SM2 with dependencies - yes, I agree. Working on it. In fact - it is the ONLY thing I have even remotely time to do now. :) A status report will arrive pretty soon on SqP.
Well, can't come up with more at this point. :)
regards, Göran
squeak-dev@lists.squeakfoundation.org
-
Colin Putney -
Diego Gomez Deck -
goran.krampe＠bluefish.se -
Michael Rueger -
stéphane ducasse