- Notifications
You must be signed in to change notification settings - Fork34
Commit1b0bc66
committed
wipe historical notes: how we created EAPI, the reasons of the design and things shaped
The pkgcore ecosystem is now on it's ~4th generation of developers and owners.This commits wipes the last real traces of the first gen, however I will commit intohistory the importance of their contributions- they were involved in getting thispig off the ground, and giving it wings. If you're reading this, it's due to no smallpart of their contributions, and I hope 20y later if you put their level ofeffort into the work, someone does similar for you.It's probably only of historical interest now- I highly doubt anyone reading thesefiles would understand the context, why things were designed this way, why EBDmattered, etc. EBD for example, much like pkgcheck's pipeline, was criticized asoverly complex, overly engineered, "astronautical architectural design", etc. Imention this as the context of what the early devs pushed through. Definitelysome of the designs flopped, andt the code execution that hasn't been rebuilt into py3kis frankly god awful since it was written compatible to py2k old style, py2k new style,and py3k. Snakeoil got it's name for a reason.EBD was built to seperate the ebuild execution state from the package manager sothat we could *strictly* run against the versioned ebuild format the ebuild adheredto. The nasty part is the state saving and reloading, a mechanism we built and thenretrofitted into portage; this is how we created EAPI, formalized into EAPI0.There's a lot of other inflections points pkgcore dev work created in gentoo; forexample, the last 2 cache formats everyone uses is ours (the git repo cache in particular),versioning mechanisms, etc. What was built generally was donated into portagewhere possible- at one point 25% of pkgcore got retrofitted into portage to replacemultiple subsystems for example. Both being helpful, but it also createdstandardization to move things forward.On to it:* Jason Stubbs- jstubbs- resolver design theory, leading to the restriction subsystem being built around CNF/DNF.* Marien Zwart- marienz. My pkgcheck pipeline design he turned into what got people caring about pkgcheck. He did a helluva lot more than that across the entire ecosystem. Folks wouldn't have cared about pkgcheck without his work to get the snowball rolling down the hill.* Marius Mauch- genone- deeply critical reviwer of pkgcore designs, specifically catching multiple intrinsic aspects of the architectural design that emergently would lead to repeating aspects of *why* I wrote this in the first placeTheir work is stil appreciated.Signed-off-by: Brian Harring <ferringb@gmail.com>1 parent7636cd8 commit1b0bc66
File tree
18 files changed
+0
-3032
lines changed- doc
- dev-notes
- formats
- framework
18 files changed
+0
-3032
lines changedThis file was deleted.
This file was deleted.
This file was deleted.
0 commit comments
Comments
(0)