On Tue, Dec 7, 2010 at 5:46 PM, Paul Davis <[email protected]> wrote: > On Tue, Dec 7, 2010 at 5:43 PM, Adam Kocoloski <[email protected]> wrote: >> On Dec 7, 2010, at 5:40 PM, Paul Davis wrote: >> >>> On Tue, Dec 7, 2010 at 5:38 PM, Adam Kocoloski <[email protected]> wrote: >>>> Hi, the mochiweb we're shipping in 1.1.0 has abandoned support for R12B05, >>>> so we should revisit our minimum required Erlang version. Do we have a >>>> compelling reason for supporting anything below R13B04? That release >>>> introduces support for recursive type specifications, which are useful >>>> when describing revision trees and JSON objects to dialyzer. >>>> >>>> Regards, Adam >>> >>> +1 for R13something. >> >> Paul, is there a NIF-based argument for a particular R13 release? I know we >> don't use NIFs in 1.1.x, but it'd be nice to limit the number of times we >> have to bump. >> >> Adam > > There's nothing major that I remember in the R13 series. Maybe a few > bug fixes or something, but I'd have to look. > > The major NIF jump was with R14. For instance, integrating Emonk requires R14. > > Also, NIF's are awesome. >
I stand corrected. Out of curiosity I went back and checked the progression of NIF support. Turns out they're not even available until R13B03. For some reason I thought the first version was in the last of the R12's. Also, in R13B04 there are some noticeable upgrades to things like NIF function signatures and other bits that would be backwards incompatible (also, no one uses the version from R13B03 anymore, so if we wanted to backport something it'd be a major breakage). So I revise my statement, I'd vote for R13B04 as the minimum. Also, it has the nice symmetry of relying on the latest R$(MAJOR)B04 Erlang VM which I declare to be the optimum balance between new features and stability.
