so... it occurs to me (very late, unfortunately), ...
# development
w
so... it occurs to me (very late, unfortunately), that it might literally have been easier to embed and ship a python interpreter that it was to get everything cross built across 6 python builds
we'd need to switch the distribution model for pants to "self contained executable"... but that's definitely a thing that we've discussed. and we now know where it could go (just in github releases)
opened https://github.com/pantsbuild/pants/issues/7369 , because i don't think we were tracking it yet.
a
i think this might also be more in line with the way we provide things like compilers
and also allows us to maybe steal ideas from facebook's xar, which doesn't need to exist
i think this is just a really really extremely good idea actually
especially because we can do whatever we want with the interpreter
w
yea.
h
This sounds cool. Good idea. Any concerns about bundle size if we’re shipping our own interpreter? I suppose with us targeting enterprise a small download size isn’t a big priority?
a
i mean, people have to download python to e.g. their docker images already, so this could replace that, which is maybe cool
i don't know what use cases would need a smaller bundle size and what that provides
i'm not against it at all i just want to know
h
Only anecdotal experience that my 20 GB Ubuntu VM I have specifically for Pants dev and nothing else frequently complains about running out of space. I think that’s much more an issue with space leaks than with our download bundle size, but has been top of my mind.
w
the OSX python installer is 27 MB. i'm guessing it's larger on Linux (since our rust binary is as well), but: not insurmountable probably.
a
i think that's because of writes to
/tmp
but like that is an intuition
h
@aloof-angle-91616 it was mostly that iirc when I used Ubuntu’s disk cleaner @witty-crayon-22786 okay 27 MB is totally reasonable IMO. Won’t post about this on issue because I don’t think that’s worth spending time fretting over. Thanks!
a
ah ok
thanks
thanks for raising this stu, getting hung up over sunk costs are what make things break down
not clear if it's definitely 100% the best idea ever but i think this is a good line of investigation
w
there is a lot to be figured out, so i'll take back how i started the thread: definitely not easier to make this switch than to cross-publish.
👍 1
but desirable nonetheless
a
maybe not easier upfront
w
yea.
r
I know I’m late to the party and everything, but I wanted to express my emphatic support for this line of investigation.