<phlebas>
i'll leave the hackmd open for the next 3 hours or so, and then send the mail (unless the others also agree with it earlier ;)
<mattip>
I would trim off any apologies like "It’s not there just yet.", and delete the middle part about "We have to add ..."
<mattip>
also, I am not sure we want HPy to be part of a shipped CPython, I would prefer it stay as a stand-alone package
<phlebas>
I was thinking that we should acknowledge somewhat the 5 year timeframe that guido is throwing about. I don't think that timeframe is even entirely unrealistic
<mattip>
moving things into CPython makes changes much more complicated
<phlebas>
of course, but eventually (when hpy is "stable enough") we would like there to be support maintained within CPython, no?
<mattip>
in my mind, no. Like pip or pytest, it can stay out of the core
<mattip>
well, eventuall if it stops changing altogether, then maybe yes. But like the timeframe, in my opinion there is no need to talk about it now
<mattip>
let's see how things evolve
<mattip>
if cython and pybind11 have hpy support next year, maybe we can push faster
<phlebas>
well, shouldn't we answer Guido's "Or am I misunderstanding the place of HPy in the ecosystem?"
<mattip>
you answered that with the first paragraph "The hope ..."
<mattip>
and with the last one "Indeed, in the long run the ..."
<phlebas>
ok, you're right
<phlebas>
Then I'll just remove that middle quote from Guido and the answer altogether
<mattip>
+1, but maybe others have different opinions
<mattip>
it is nice that Guido approves of the approach
<phlebas>
yes
<steve_s>
I wonder if mentioning Numpy, Cython and maybe pybind11 (though we haven't started coding anything there yet) can add some credibility/sign of seriousness of the effort
vstinner has joined #hpy
<antocuni>
phlebas: the email looks good to me
<antocuni>
mattip: FWIW, I'd like HPy to be officially part of CPython eventually. But I agree it cannot happen soon
Techcable has quit [Ping timeout: 265 seconds]
<phlebas>
ok, sent
<phlebas>
mattip: I tried sending twice, using an email program and the mailman web interface. Maybe it's blocking @oracle.com mailing address ;) :P
<phlebas>
Could you send the email since your mails do seem to be getting through? I've used exactly the text as it's in the HackMD right now
Techcable has joined #hpy
<vstinner>
phlebas: i'm now curious about that email :-)
<vstinner>
phlebas: i'm enjoying the discussion about the C API on python-dev :)
<phlebas>
i tried sending it via my mail client and the mailman web interface - nope, not showing. Marc-Andre Lemburg checked, and it's not held for moderation, so ... ¯\_(ツ)_/¯
<phlebas>
yes, that's what I used too ... well, no idea
<mattip>
I will send it in the name of the team
<phlebas>
thank you :)
<vstinner>
i'm now trying to write down articles about incompatible changes in Python, how it went previously, what can be done to reduce mitigate compatibility issues, etc.
<vstinner>
it's a hot topic :-)
<phlebas>
vstinner: I'm really glad you shared those tools for checking the top packages. as the responses showed, there is more to be done, but it's an important step to start looking at the impact for projects on a larger scale in this manner
<vstinner>
phlebas: is there a plan to update the hpy repo to top 4000 projects?
<vstinner>
you cannot imagine how many projects contain random binary files, like.. a full virtual environment containing not 1 but *3* python ELF binary files!
<vstinner>
the script searchs in any file, Python or C code, and uses a regex. so it can be used to search for C code
<mattip>
done. I never understand how to thread the reply in the right place on these long discussions, I hope it is OK
<phlebas>
mattip: thank you. for whatever reason one of my emails just got through. must have been some blockage in the email pipes that lead from the oracle mail server to the mailing list :(
<mattip>
cool
phlebas has quit [Read error: Connection reset by peer]