Hi Lex!
"Lex Spoon" lex@cc.gatech.edu wrote:
The general release plan sounds great: clean up 3.8/3.9, then for 4.0 focus on partitioning.
Mmmm, that is not exactly what the report says though. :)
We wrote that the partitioning work (staking out the image using PI (or similar) and assigning parts to maintainers) should start ASAP in the 3.9 cycle, not in 4.0. This is one of the bits that the Packages Team will do IMHO. That Team isn't formally formed yet because the Team leader question is still open - but I did create the mailinglist yesterday:
packages-subscribe@discuss.squeakfoundation.org
But the Morphic split will not happen until 4.0, that is true.
(please. PLEASE. stop speaking in codewords like "TFNR". It causes people who might want to help, to click "delete" because they don't even know what the message is about.)
Sure, I can stop using that acronym, no worries. I just thought most of us do know what it is.
However, two important items were left off the list:
- Work out a way to release a set of packages this time, and not just
an image. If this is nothing else than a zip file with a bunch of packages in a subdirectory, that would be fine. We really need this to give people a stable basis for working in Squeak -- especially as more and more stuff ends up in packages.
Sounds reasonable. I assume you and I (and the rest) will discuss it further on the Packages Team.
- Get a bug tracker going so that we can see what specific packages
have open release-critical bugs in them. Otherwise, the work in #1 is much less useful, because we have no way to tell how stable different packages are.
For the "official packages" and for the staked out partitions of the image I simply think we should use Mantis just like we already do. We already use it for the image and for some packages.
But the specifics for 3.9 in this regard is for the Janitors Team formed by Ken Causey.
Additionally, the UI abstraction task is fine but seems relatively unimportant compared to these other items. How muh code, really, is going to find it useful to target 5 different UI's? Some, to be sure... but the above two items are easy and important. Without them, dividing the image into packages can actually *hurt* people who want to use Squeak (instead of developing it further).
Just a note - the partitioning effort is not primarily focused on actually turning those parts into packages (that are external to the image). This I think you are aware of.
And getting ToolBuilder into 3.9 is an important piece of the puzzle for 4.0 - as I think the report explained. It will not be primarily important for 3.9 itself.
For motivation, just picture someone who wants to develop software in Squeak, but not take part in developing Squeak itself. They decide they want to try Monticello... boom. Why boom, when it's available on SqueakMap? Because they are using a 3.5 image, because 3.6 and 3.7 broke their networking code, and because 3.8 and 3.9 seem too unstable, and because people keep posting new packages to 3.5 without thoroughly testing them. It should be normal that people can actually use stable releases to develop in, but doing so is a hassle if both (a) everything is in packages and (b) there are no stable releases of packages.
This we all agree on I think.
Lex
regards, Göran