You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Since these conversations on slack apparently got lost in the enthusiasm of refactoring, I thought I'd write down my reasoning behind asking for a 2.1 release before we merge in the major Cython refactoring in #1595.
Cython is going to break some installs in unpredictable ways.
The current master is pretty well tested and vetted, and it has bug fixes and performance improvements that will make users' lives better without breaking anything (due to semantic versioning requirements).
The first two reasons combined imply that we would incur some significant support burden from releasing a Cython-based version, whereas we would incur little to no additional support burden from releasing the current master as khmer 2.1.
I predict that there will be some reasonably annoying and significant bugs and compilation challenges with the Cython branch that will cost us dev time and slow down any Cython-based release. Best to let the Cython branch delay development and release of 3.0 with a well-tested 2.1 update out there in the wild.
A sourmash 2.0 release depends on a khmer 2.1, and sourmash 2.0 should be ready soon-ish.
It looks like the transition to oxli is well under way in Split CPython and begin the Cython revolution #1595 so we can plan a bunch of changes with 3.0, rather than pushing out a 3.0 sooner and confusing people with a 4.0. (Not sure how this will actually work out, but it seems reasonable).
So, in sum, I'm cautiously optimistic about Cython but there is enough potential breakage in the refactoring that I want to get khmer 2.1 out first.
The text was updated successfully, but these errors were encountered:
Incidentally, #1623 is another great example of why we do want to get 2.1 out soon, so that we can add exciting new command line features and break backwards compatibility of scripts, on top of all the cool let's-build-on-khmer stuff that's motivating @standage and @camillescott's. So I'm enthusiastic about getting 2.1 out. But it's not something I can move forward on my own until April sometime.
Since these conversations on slack apparently got lost in the enthusiasm of refactoring, I thought I'd write down my reasoning behind asking for a 2.1 release before we merge in the major Cython refactoring in #1595.
Cython is going to break some installs in unpredictable ways.
The current master is pretty well tested and vetted, and it has bug fixes and performance improvements that will make users' lives better without breaking anything (due to semantic versioning requirements).
The first two reasons combined imply that we would incur some significant support burden from releasing a Cython-based version, whereas we would incur little to no additional support burden from releasing the current master as khmer 2.1.
I predict that there will be some reasonably annoying and significant bugs and compilation challenges with the Cython branch that will cost us dev time and slow down any Cython-based release. Best to let the Cython branch delay development and release of 3.0 with a well-tested 2.1 update out there in the wild.
A sourmash 2.0 release depends on a khmer 2.1, and sourmash 2.0 should be ready soon-ish.
It looks like the transition to oxli is well under way in Split CPython and begin the Cython revolution #1595 so we can plan a bunch of changes with 3.0, rather than pushing out a 3.0 sooner and confusing people with a 4.0. (Not sure how this will actually work out, but it seems reasonable).
So, in sum, I'm cautiously optimistic about Cython but there is enough potential breakage in the refactoring that I want to get khmer 2.1 out first.
The text was updated successfully, but these errors were encountered: