Log file Viewer
#python2.8 logs for Wednesday the 22nd of January, 2014
[13:31:26] <mitsuhiko> i'm just ignoring the existence of python 3 and i immediately feel better about myself :)
[13:32:07] <tos9> yeah, it's unfortunate that the crux of the issue here is that most of the people who could probably make things better have no interest in doing so because of how well Python 2 works :)
[13:32:30] <tos9> (or that "making things less borked is hard", one of the two, still can't tell which)
[13:36:56] <regebro> Well, if somebody could tell the core devs what "better" meant, this would be a step forward.
[13:37:44] <regebro> There is a lot of people caling for Python 2.8, but I can't see many people with opinions on what it would entail.
[16:20:26] <dash> regebro: all of the features that were included in 3.x that could be added to 2.x compatibly but weren't
[16:22:36] <dash> if the only goal is to drive adoption of 3.x, then there isn't any conversation to have
[16:24:09] <dash> since it's clear that the developers of python don't share goals with a large segment of 2.x users
[16:25:07] <dash> a lot of the confusion and argument has come from misunderstanding what the python devs really want
[16:25:26] <regebro> The problem is that people find it difficult or too much work to move to Python 3.
[16:27:00] <regebro> Well, it's only a problem in as much as Zope 2 isn't likely to run on Python 3.
[16:27:43] <regebro> But there really isn't any version of Python that can fix that, Zope 2 has so much heavy Python magic, that even moving from Python 2.4 was a major effort that too a long time.
[16:28:53] <regebro> Other than that I tend to use Python 3, so this is clearly not a problem for me.
[16:29:49] <regebro> In other words, if the problem isn't what people claim the problem is, then I don't know what the problem is.
[16:32:13] <regebro> dash: The complaints on various blogs etc is that it's hard to move to Python 3. The suggested solution has been Python 2.8.
[16:32:34] <regebro> I don't understand what that 2.8 would entail, so I came here to see if anyone had any suggestions.
[16:32:48] <regebro> You then claim that the claimed problem is not the problem. So what is the problem?
[16:34:20] <dash> -- other people just want improved facilities available to their python programs right now
[16:35:52] <regebro> So, your problem is that Python 3 is better than Python 2, and you want Python 2.8 to fix that, to make Python 2 (almost) as good as Python 3?
[16:38:26] <regebro> So Python 3 is not better than Python 2, despite having benefits, because it's not Python 2?
[16:39:09] <dstufft> I believe that Python 3 is a better language in general, but for some projects the benefit isn't that large and the porting effort is much larger then the benefits would reward
[16:39:54] <dstufft> I also think that backporting things to 2.8 like yield from would effectively kill Python3 (which is probably something dash would prefer)
[16:40:17] <regebro> I'm just arguing against dash saying that Python 3 is not better, because although it's better it's not Python 2.
[16:41:08] <regebro> dstufft: I don't know if it would kill it, but it would probably slow down adoption. And it seems that most people agree that the problem is that it's too hard to Port to Python 3.
[16:42:03] <tos9> regebro: dash's argument is mostly "python 2 is going to stay, do you care about that [and are willing to help those who will need or want to continue writing it] or not?
[16:42:31] <dash> another cost is that if you want to use python 3 for your project you have to pick between complicating your codebase with polyglot bits, or reducing your deployment options
[16:42:59] <regebro> I do agree that if you assume Python 3 is a failure, then Python 2.8 makes sense, and its' fairly straightforward what to include.
[16:43:01] <tos9> Personally, I don't, not that much :D, even though I am generally someone who writes Python2 almost exclusively, so I do want things to help Python 3 along, and don't think 2.8 is going to help, but I *do* think that adding any and everything reasonable to Python 3 that will help people writing cross compatible code should be on the table for Py3
[16:44:34] <tos9> Twisted and PyPy are still the major 2 things missing [certainly for me/us], not counting all of the thigns that are really slow on Py3 that we use.
[16:45:24] <dstufft> tos9: I agree fwiw, that adding everything reasonable to Py3 to help porting is the way forward
[16:45:42] <dstufft> tos9: tbh that should have been the case from day 1, it sucks it wasn't, but it can be fixed going forward
[16:46:35] <dstufft> I'm working on rewriting PyPI, I started with Python3, and every other library needed ported
[16:46:44] <tos9> dstufft: There's still a bit too much pushback on certain things I think. It's hard to discern whether it's actual sane pushback or if it's just the usual mailing list insanities.
[16:47:26] <dstufft> but you're not likely to see a lot of real world Python3 use until you can be reasonably sure that the modules you want to use from PyPI are going to work on py3
[16:47:51] <tos9> dstufft: Well I know *your* opinion given this morning's antics :D [not that I disagree]
[16:49:08] <tos9> Someone in #python brought up PEP 3113 a couple of days ago again, another thing that should just die a swift death.
[16:51:28] <dash> regebro: the joke? well. do you believe python 3 has fewer modules/libraries available than python 2? more specifically, that there are many python 2 libraries not available on python 3?
[16:52:16] <regebro> dash: There are thousands of modules on PyPI that is used only by it's author, and uploaded once and never again. The answer is "yes".
[16:53:20] <regebro> The question was which libraries you are missing. Requiring that all modules on PyPI run on Python 3 is not constructive.
[17:01:32] <dash> there are some benefits with python 3, naturally. some of those benefits even come from breaking compatibility
[17:03:54] <regebro> dash: So you *do* agree that Python 3 is a better language? As the drawbacks you mention is essentially that it's incompatible with Python 2.
[17:05:25] <regebro> Which returns us back half an hour, in which I ask you if your problem is that Python 3 is better than Python 2.
[17:06:15] <regebro> OK, I don't think we are getting anywhere on that line of discussion, it seems to have stalled.
[17:08:31] <dash> if i'm going to pick a language for a new project, or to port old code to, python 3 isn't my only choice
[17:09:38] <regebro> dash: So your problem is that you don't want to use Python 3, and you are annoyed that development is on Python 3.
[17:10:44] <regebro> Most people who express an interest in Python 2.8 or otherways complain about Python 3 are interested in using it.
[17:14:39] <regebro> Because there are changes and improvements that can not be made in a compatible way.
[17:15:46] <regebro> Hence, your problem with Python 3 is that it's better than Python 2, in non-compatible ways.
[17:17:14] <dash> my problem with python 3 is that its promoters treat it as morally superior and a historical inevitability
[17:19:11] <regebro> dash: No, they don't treat it as a "historical inevitability". They treat it as superior (because it is).
[17:19:41] <regebro> And they also treat it A Good Thing that they want to use and they want others to use.
[17:20:59] <regebro> dash: In what way can they take compatibility more seriously than they already do?
[17:21:49] <dash> regebro: pretty obvious that being able to run python 2.7/python 2.8 programs unmodified on python 3.x would be a start
[17:22:17] <regebro> Is your problem with Python 3 that it's being developed? Because then then this discussion is really pointless.
[17:23:44] <wangofett> "we've been running Python 2.7 programs all along, therefore Python 2.7 is better"
[17:24:09] <dstufft> wangofett: you're thinking in abstract, dash is thinking in terms of praticality
[17:24:13] <regebro> If we make Python 3.5 run Python 2 programs unmodified, then all Python 3 code would break, etc.
[17:24:22] <dash> wangofett: shrug, it's a reason to use python 2, and it's not a reason to use python 3
[17:24:47] <wangofett> you can make the argument that it's better to not rewrite code than it is to rewrite code if it works already.
[17:24:55] <dash> regebro: people have figured out ways to make incompatible things work together in the past
[17:25:29] <dash> regebro: if you start from the belief that this is "NOT POSSIBLE", then obviously no way to do it will be found
[17:25:30] <regebro> dash: Yes, and there are many things in Python 2.7 to help you overcome the incompatibilities.
[17:26:13] <wangofett> unless your only qualification for better is that it runs your code right now. In which case, say so.
[17:26:40] <dash> wangofett: has it occurred to you that things can be good or bad in more than one way? :)
[17:27:32] <dash> to be clear: I have been using python as my primary development environment for 15 years
[17:27:55] <wangofett> dash: sure. But if you're arguing that Python 2 is better because it runs code right now, and Python 3 is worse because people are trying to push you to use Python 3 that does not make any sense.
[17:29:06] <regebro> dash: Yes, the aim is, that as with all older Python versions, Python 2 will not be supported in some future.
[17:29:52] <dash> I am being told that python 2.x will no longer be supported in the future, *and that I need to rewrite my code if I want to use a language that is supported*.
[17:30:07] <regebro> dash: Yes, the aim is, that as with all older Python versions, Python 2 will not be supported in some future.
[17:30:12] <dash> regebro: So apparently it's time for me to pick a new language -- you want me to pick python 3
[17:30:15] <dstufft> The text model switch is something that I don't believe could have been introduced in a backwards compatible way
[17:30:23] <tomprince> python 1.x and python 2.x are essential different version of the same language, python 3.x is fundamentally a *different* language.
[17:30:56] <dash> wangofett: python 2.x ran python 1.x programs, python 3.x does not run python 2.x programs
[17:31:22] <regebro> I do think however, that when you compare moving from Python 2 to Python 3 with moving from Python to well, whetever else, you are wrong.
[17:32:11] <dash> I have to choose between a platform that doesn't run my existing python code, and some other platform that doesn't run my existing python code
[17:32:47] <regebro> Moving to Python 3 is not comparable in cost/work with rewriting something in a different language.
[17:32:49] <dstufft> dash: that's being somewhat disingenuous, Porting to Python3 is very rarely anything like a complete rewrite. ;)
[17:33:46] <dstufft> dash: I don't know, but I suspect that Twisted wasn't completely rewritten for Python3 in the parts that were ported ;)
[17:34:01] <dash> to put it another way: over the last 10 years, whenever I have needed to pick a language for a new project, python has gotten a bonus in that decision making process because it makes interoperating with my existing stuff easier
[17:34:54] <dstufft> Python3 can be a signifcant amount of effort to port to depending on what you're doing though, and it can make a lot of sense to switch to another language if you're going to have to expend singifncant effort and expend some more effort for different benefits
[17:35:55] <tomprince> dstufft: It wasn't a complete rewrite. That doesn't mean that it was less effort than a complete rewrite.
[17:36:44] <regebro> And the effort to move to Python 2 gets multiplied if you are using doctests for all tests for example.
[17:39:15] <dstufft> Also porting to Python3 is effiectively impossible if you don't have unit tests
[17:40:31] <dash> Anyway, it's pretty obvious that python 2 programs could be supported on python 3, in some fashion, if someone really wanted them to be
[17:42:21] <dash> regebro: people translate code in one language to another all the time, it's called a "compiler"
[17:43:14] <dstufft> dash: text/bytes split on Python3 requires some decision making process since there's no automatic method of determning if bytes or str is the correct type for some bit of code
[17:43:44] <regebro> If you mean that it would be possible to use the same executable for both Python 2 and Python 3, dependning on for example what name it was called with, then yes.
[17:44:49] <regebro> Well, 2to3 does most of that already. But there differences both large and small, that make it hard.
[17:45:13] <dash> the unclever solution would be to compile code creating 'str' and 'unicode' objects to code that created 'python2str' and 'python2unicode' objects
[17:46:35] <_habnabit> regebro, like i said, 2to3 does practically nothing. you still need version checks for a number of trivial operations
[17:48:27] <dash> regebro: how does clojurescript/brython/coffeescript/etc get compiled to js? how does rpython get compiled to C
[17:49:52] <dash> regebro: i already wrote a compiler but it's for a language I like better than python
[17:50:49] <dash> regebro: anyway consider than pypy seems quite capable of maintaing python3 and python2 support in the same codebase
[17:51:53] <dstufft> regebro: I think the misunderstanding here is that a lot of the current "hey port to py3" efforts involve effort, sometimes a lot of it, on the people whose stuff got broken by the core developers. So it's a bit like knocking a cup out of someones hand and then telling them they should clean it up
[17:52:52] <tomprince> Well, it is a seperate python program. It could certainly be turned into a library.
[17:52:56] <regebro> dstufft: I completely understand that. I came here because loads of people seem to think that Python 2.8 would make things better.
[17:53:35] <dstufft> regebro: I think that most people who want python 2.8 just don't want to do the extra work of porting to Python3, and that's totally a legit view point
[17:54:16] <regebro> I still haven't seen any constructive suggestions on how 2.8 could make porting easier anyway.
[17:54:59] <dash> chained exceptions is a nice-to-have. but that's not even an incompatible change afaict
[17:55:11] <dstufft> regebro: I've seen one, that would enable the breaking changes that python3 did, but enable you to turn them on one at a time
[17:55:56] <tomprince> regebro: Well, most of the people who are active at this minute don't have a desire to port, but don't see why that should mean they don't get access to the compatible improvements of python3.
[17:56:41] <tomprince> Not everybody who comes here shares that view, but most of those people seem to be AFK.
[17:56:46] <dstufft> tomprince: well they shouldn't get access to the compatible improvements because nobody's been motivated enough to maintain a python 2.x branch that has them and the python-dev folks aren't mtivated to do that either :)
[17:58:20] <tomprince> Hmmm, i wondr if pypy could be made smart enough to detect the usage pattern that leads to 'yield from' and automatically convert it.
[17:58:23] <dstufft> tomprince: demanding that python-dev maintain 2.x when they don't want to is akin to demanding that people port to python3, except python2 will continue working so nobodies forcing py2 users to switch if they don't want to (aside from people who want to try, but that's an impossible situation anyways)
[17:58:31] <regebro> tomprince: In other words, you are with dash on the topic: You simply don't want to port. I'm less interested and supportive of that standpoint that I was a couple of hours ago.
[17:59:25] <regebro> Right, I mean, people who don't want to port, but still demand all the benefits of Python 3.
[17:59:57] <tomprince> dstufft: Sure. But it sometimes seems like (some of) python-dev et al expect our labour to port to python3.
[18:00:21] <dash> regebro: and this is why i started by saying everything 2.x-related should just go ahead and be removed
[18:00:46] <dstufft> I don't think dash is demanding them, he's just stated that Python3 isn't interesting to him and he'd rather expend his efforts either maintaing his Python2 code or working on another language that offers more benefits to him
[18:01:06] <dstufft> and saying that he'd like a 2.8 that had new features (who doesn't like new features?)
[18:01:13] <dash> regebro: all I mean is that the python developers have made it clear they don't want users who aren't interested in 3.x
[18:01:42] <regebro> dash: Nobody want to *force* anyone to Python 3. No-one is interested in removing any Python 2.x-related stuff.
[18:01:54] <dstufft> tomprince: the reasoning for pip not being in 2.7 was some bullshit about it might break something for sombody
[18:02:04] <regebro> We are interested in helping people that want to be on Python 3 to be able to get onto Python 3.
[18:02:11] <dstufft> and there's no 2.8 because python-dev isn't interested in maintaing 2.x anymore
[18:02:40] <regebro> Attitudes like "Well then just remove all Python 2 related things" just makes people *less* interested in helping.
[18:02:41] <dstufft> lots of people want to force people on Py3, (lots of people want to force Py3 to be abandoned and switch back to Py2 too)
[18:03:13] <dash> dstufft: well, people in the latter camp don't understand the open source software lifecycle
[18:03:24] <regebro> tomprince: Which is completely different from removing Python 2 from python.org, fo ex.
[18:03:56] <regebro> tomprince: Python-dev is also not really "blocking" anything. They are just not interested in improving it.
[18:05:15] <dstufft> I assume if they were going to be releasing a 2.8 they would have allowed me to add it in that
[18:05:59] <dstufft> I think not including pip is a user hostile move when you had someone willing to do all the work, but it's not super cut and dry
[18:06:01] <tomprince> regebro: So only bugs can be added to python 2.x because python-dev doesn't want to work on it? Even if somebody steps up to do the work? That still sounds like blocking.
[18:07:11] <dstufft> tomprince: theortically only bugs can be added to 2.7.x because that's the backwards compat policy
[18:07:15] <regebro> tomprince: They are blocking this "somebody" from forking Python 2 and releasing it from Python 2.8, yes.
[18:07:51] <dstufft> it just so happens it works out to the fact that 2.7 is also the last 2.x so it's effectively preventing new features from 2.x
[18:08:15] <dstufft> but adding new features to 2.x would require either relaxing the backwards compat policy, or someone to do the work for a 2.8
[18:09:02] <regebro> tomprince: If you find enough developers that are interested in maintaining Python 2.x, then that situation might change.
[18:09:40] <regebro> But the decision amongst the current maintainers is that they were not interested in making a Python 2.8.
[18:11:06] <regebro> tomprince: You can call that "blocking" if you want, but it's stretching it, I think. Especially since nobody is being blocked. :-)
[18:12:47] <regebro> But in any case, none of this is what I came here to discuss, so I should stop now. :-)
[18:23:30] <eriknw> I understand the desire of devs to want to focus on python 3, but the big caveat is that there isn't a substantial userbase using python3
[19:39:07] <eriknw> a substantial userbase gradually moved between python 2.0 to 2.7, and there are significant differences between python 2 versions
[19:41:36] <eriknw> for many (not all) things, updating from python 2.7 to python 3.x isn't so different from updating to a newer python 2 version
[19:43:05] <eriknw> there was already an upgrade path that used mostly small increments. the path between python 2 and python 3 is a large increment. I guess the idea behind python 2.8 would be to again make the increments smaller
[20:03:32] <regebro> eriknw: They have always concentrated on the latest version of Python. There has always been gradual movement of userbase.
[20:06:26] <eriknw> regebro, a list of things that can go on a wiki that can later be discussed and modified
[20:07:33] <eriknw> right now there isn't a clear list of what python 2.8 might entail. as with previous python versions, it should also give deprecation warnings instead of just being deprecated (as in python 3)
[20:09:31] <regebro> eriknw: I've been asking for things to go into 2.8 for years by now, nobody can come up with anything much. So "a list of things" is not an answer I find particularily useful.
[20:10:07] <regebro> eriknw: It's true that there aren't deprecation warnings for things that are fixed by 2to3.
[20:10:40] <regebro> Any of course there might be oversights. I'm not sure this really is a big issue though.
[20:11:36] <dstufft> I don't even know what a yield from does because it's not in 2.x and thus useless to me since I can't py3 only yet :[
[20:11:52] <tos9> exception chaining, if we're naming features that are good about py3 that are going unused because they're py3-only
[20:11:54] <tomprince> An option for clear seperation between bytes and unicode (i.e. turning off automatic conversion)
[20:11:55] <regebro> tomprince: We are now talking about things to make porting easier, not things to help you avoid Python 3.
[20:12:38] <regebro> eriknw: Now you got a huge plus point, because you are literally the first one to actually come up with something. :-)
[20:13:06] <regebro> I have some ideas what could go in there, but I refuse to tell anyone, as I want to know what people would find useful.
[20:13:38] <tos9> regebro: it's still unclear to me whether you realize there are 2 groups asking for things
[20:15:49] <regebro> Anyway, backporting new features belongs to "helping me avoid Python 3" which I don't thing is the right way to go.
[20:16:28] <dstufft> "Give my Python 2.8 I don't want 3.x at all", "Give me 2.8 to ease my porting to 3.x", "Give me 3.x to ease my porting to 3.x"
[20:17:46] <regebro> Right, and I don't think the first is interesting unless Python 3 is declared a failure.
[20:17:48] <tos9> I was dividing them functionally between "we are being neglected" and "we want to port whether or not we like Py3 but can't yet easily"
[20:18:53] <regebro> eriknw: To clarify: lambdas are anonymous functions. I'm not aware of any other was to create anonymous functions (except for creating function objects directly, but...)
[20:19:36] <dstufft> tos9: they deserve attention, but there's not any meaningful thing for python-dev to do unless python-dev recinds PEP404
[20:19:58] <regebro> tos9: Again, nobody is forcing them. It's a valid opinion to claim that Python 3 was a mistake, but currently the Python developers do not agree.
[20:20:26] <tos9> regebro: Please separate "we don't like Python 3" which is a sentiment this group may feel from "hey we want to write programs do you want to help us or not"
[20:20:58] <regebro> So I'm not interested in discussing how to help people avoid Python 3, because that's not going to happen unless Python 3 is declared a failure, which has not happened.
[20:21:43] <dstufft> also plenty of people don't like Python3 but are working to port their things to Python3 anyways
[20:22:19] <regebro> So we have now one suggestion of what Python 2.8 could include. That's good, but it's not really that much. :-)
[20:22:29] <tos9> Sorry, I can't put that any more clearly I think :), whether or not it's up for discussion, it's a fact, there is such a group and they are partially the ones who want a 2.8. If you want to discuss 2.8 you have to include them. If you are just answering "no I don't care about any of you too bad" then that's fine (wrong IMHO even if 2.8 isn't a thing, but fine)
[20:23:06] <tos9> At that point the question becomes "will 2.8 help people port to Py3" -- as I said I personally don't think so but I also am not too bothered by the channel name, and more by the sentiment which I do agree with
[20:23:43] <regebro> eriknw: Also, the same effect can be largely gotten by `six` and `futurize` or whatever it was called. Both libraries help you write code that will run under Python 2 and Python 3.
[20:24:51] <eriknw> yeah, regebro, more people are creating libraries compatible with python 2 and 3 (myself included)
[20:29:18] <eriknw> heck, instead of introducing python 2.8, perhaps the next version of 2.7 could have more things added to __future__
[20:34:43] <eriknw> then perhaps someone could make a "__python3__" library that allows users to use it like "__future__", such as: "from __python3__ import whatever"
[20:37:56] <richmoore2> yes, but having a python 2.8 that provides them in a consistent way would give a stable platform to work with
[20:42:12] <tos9> you declare some stuff. your user installs your thing and all of its deps. Packaging ain't perfect, but I don't think we need 2.8 to download a bunch of tarballs :)
[20:43:23] <richmoore2> i want to be able to get stuff working so that people can simply use it once it's installed. virtualenv doesn't help there
[20:44:22] <richmoore2> a python interpreter with 2.7 features and updated libraries would be a massive win for people like me (though I'm sure it would annoy people who think we should move to 3.x)
[20:55:35] <richmoore2> that still prevents you doing proper package management signature verification etc.
[20:56:49] <richmoore2> i'm pretty happy with 2.7 (sure it has warts) but i really want some of the newer libraries to be shipped as official components
[20:58:18] <tomprince> richmoore2: It sounds like you would benefit from better intregration of python and distro packaging tools, than from a py2.8.
[21:00:00] <richmoore2> tomprince: if there was a 2.8 that said "this is the updated standard library" then that would get packaged
[21:03:26] <tomprince> I'm just suggesting that and updated standard library isn't really what you want.
[21:03:47] <richmoore2> point in case would be the embedded urllib3 in requests rather than fixing urllib2 (or urllib)
[21:04:29] <richmoore2> perhaps what i'd really like is a python 2.7 expansion pack - a set of libraries
[21:37:03] <regebro> What we need, and what I'm arguing for, is a list of things needed to make porting easier/more gradual.
[21:39:03] <eriknw> how do you propose that list get created? this channel isn't particularly useful for creating retrievable knowledge
[21:53:10] <eriknw> yes, aclark, although it's hardly a productive platform for creating and maintaining a list
[21:55:07] <eriknw> chatting on this channel requires humans to do work too. hasn't seemed particularly productive from what I've seen
[21:55:42] <eriknw> no offense to those who chat here. based on the number of people in this channel, there is obviously interest in this topic
[21:57:25] <aclark> i think the fact that 1.) it exists and 2.) there are people in here is impressive, regardless of whether or not there is ever a 2.8
[22:01:42] <aclark> And there are compelling arguments on both sides e.g. http://alexgaynor.net/2013/dec/30/about-python-3/ vs. http://ncoghlan-devs-python-notes.readthedocs.org/en/latest/python3/index.html
[22:12:11] <richmoore2> unicode is really not an issue for my use of python, so for me afaics the only benefit is incremental improvements to the library (which i think should be backported) and yield from
[22:15:22] <eriknw> it seems that most "users" of python 3 are library developers who support both python2.6+ and python 3.3+, which results in the code being written using the common denominator. I know very few python3-only projects (although I'm sure many more exist than I know about)
[22:17:37] <eriknw> unless you need to handle unicode as done in python3, I don't see much incentive to use 3
[22:30:33] <regebro> eriknw: Considering that we so far has exactly 1 proposal, maintaining the list is not a big problem. :-)
[22:32:54] <richmoore2> ie do most people actually care about porting to 3.x or do they just want 2.x to be improved.
[22:34:03] <regebro> richmoore2: As long as Python 3 is alive, it's not going to happen, so it's not relevant.
[22:34:47] <regebro> Unless of course somebody forks Python 2, and makes a release under another name, and then everybody just moves to that, showing that the Python Core developers are all wrong.
[22:35:02] <richmoore2> and i think a defacto update to the libraries is going to happen. the only question is when
[22:37:03] <richmoore2> i mean that linux distros are making available backported libraries in a convenient way (including those that were originally targeted at 3.x only) meaning that the library is evolving even without the core python team
[22:37:47] <regebro> richmoore2: not only linux distros, several of the new libraries are available on PyPI.
[22:41:46] <richmoore2> regebro: it's harder for sysadmins and packagers when you use 3rd party installs. it also doesn't work when you're using shared readonly file systems for /usr etc.
[22:42:19] <regebro> richmoore2: You just pointed out that the linux distros more and more are providing these.
[22:42:20] <richmoore2> regebro: you also can't verify the package signatures to ensure there's nothing malicious in there
[22:43:11] <richmoore2> regebro: as i said, this is starting to happen in a defacto way, but i'd like an official baseline - "you can rely on these being present"
[22:44:54] <regebro> OK. Well, in any case, this is still in the realm of "how do I avoid moving to Python 3", so it's still IMO not relevant at the moment.
[22:45:55] <richmoore2> i don't see any reasons why i would want to use python3. i've read the various articles etc. and it offers me little
[22:46:06] <regebro> Once again: As long as Python 3 is not deemed a dead end, making people avoid using Python 3 is not going to be a high priority for the core developers.
[22:46:43] <regebro> Hence, a Python 2.8 whose aim is to make people not move to Python 3 simply is not likely to happen.
[22:46:47] <richmoore2> i'm not saying it should be. but i don't care about the core language - that works as is for me
[22:47:39] <regebro> We therefore end up discussing what hypothetically could go into a hypothetical version.
[22:48:10] <regebro> Should Python 3 be declared a dead end, what will go into Python 2.8 is rather more obvious, and hence do not really need discussing.
[22:48:10] <richmoore2> yes, and i'd hoped this channel was going to be about how to turn it into something that wasn't hypothetical
[22:48:11] <eriknw> the goal of 2.8 isn't to make people not move to Python 3. That's just dense. It's to provide additional backwards compatible features to the vast majority of python users (who happen to still use python 2)
[22:48:26] <regebro> richmoore2: Therefore, not discussing this is not wrongheaded. Discussing it is pointless.
[22:48:59] <richmoore2> for me the aim of a python 2.8 would be to improve the development facilities for people using python < 3
[22:49:14] <regebro> eriknw: As long as Python 3 is not deaclared a dead end, helping people move to Python 3 is indeed the only reason for a Python 2.8.
[22:49:43] <richmoore2> that could (and possibly should) include improved facilities for those who want to move, but should also offer things for those who don't
[22:49:46] <regebro> richmoore2: Once again: That will not happen, unless you fork Python 2 and do it yourself, or Python 3 is declared a dead end.
[22:50:37] <eriknw> that (1) the point of 2.8 is to help people move to python 3, and (2) 2.8 is to make people NOT move to python 3
[22:50:44] <richmoore2> regebro: there's no need for me to do that. simply providing a package of a bunch of the available libraries would actually meet my goal
[22:51:27] <eriknw> <regebro> Hence, a Python 2.8 whose aim is to make people not move to Python 3 simply is not likely to happen.
[22:53:47] <eriknw> goodness forbid that developmental effort be given to the language the vast majority of users still use
[22:54:39] <richmoore2> in my experience with libraries like Qt the people who write the library aren't the people who write the apps, and i suspect that's the same here
[22:54:43] <eriknw> if python 2.8 makes python 2 more palatable for another few years, which slows down adoption of python 3, then who cares? what's the problem with that?
[22:55:59] <dash> eriknw: or makes python 2 more palatable for another few years, increasing adoption of python 3... :)
[22:56:01] <regebro> eriknw: It will enable everyone to continue to complain about how no-one is using Python 3 for another few years.
[22:56:11] <richmoore2> in the case of python if the people who write the language aren't the ones wrote actually write the library (cf the comment above from someone about the standard library being where classes go to die) then maybe decoupling the two is what's needed
[22:56:34] <eriknw> regebro, so what? it will help *users* use the language they are currently using and like
[22:56:35] <regebro> Which means we'll just have this discussion again about Python 2.9 in five years, etc, etc.
[22:57:08] <regebro> No, I realize that for people who don't want to use Python 3 it is not a problem that nobody uses Python 3.
[22:57:38] <regebro> I also realize that it's going to be impossible to convince those people that moving to Python 3 is a good thing.
[22:57:54] <richmoore2> you're assigning desire where it's not warranted. i don't /care/ about python 3. i don't have any hatred for it
[22:58:25] <dstufft> Obviously the idea is that people will be forced to move to Python3 by Python2 EOLing, which is silly
[22:58:53] <richmoore2> iirc (including rereading the topic) this is the place for exactly this discussion
[23:00:54] <richmoore2> but you're sure we're wrong? (thinking of your comments about it being impossible to convince us?)
[23:01:45] <regebro> This is not a discussion issue. It's better. Full stop. If you see it or not makes no difference.
[23:02:13] <regebro> I agree that the cost of moving from Python 2 can be large. Therefore the discussion should be of how to lower that cost.
[23:02:32] <eriknw> it's often not a personal decision. many companies have developed an extensive internal code-base using python2, and it is very difficult to compel a change that isn't required and has limited payoff
[23:02:35] <richmoore2> why? surely the discussion of if the improvement is worth it is equally valid?
[23:03:05] <dash> regebro: well, that's all we _can_ talk about, since the time period where we could talk about increasing the benefits is over
[23:03:27] <regebro> If you think that the current cost sin't worth it, then it is not worth it. That's a tautology.
[23:04:30] <regebro> dash: No, but increasing the benefits, ie improving Python 3, is something that is discussed on Python-dev and Python-ideas already.
[23:05:01] <richmoore2> that's certainly one option. however in the interests of balance you must admit that the other option is that if the main win from python 3 is the library improvements which can and have been backported then a release that includes those improvements is a valid thing to do
[23:05:08] <dash> regebro: mmm, i thought we were agreed that the crucial benefits of python 3 were the ones resulting from breaking compatibility
[23:05:32] <eriknw> lowering the cost sufficient for quick adoption by more companies I think would require some degree of python2 compatibility
[23:05:56] <regebro> dash: No, it is Python 3, because it breaks compatibility. If it didn't do that it would be called Python 2.
[23:06:17] <dash> richmoore2: I think that improving the experience of python 2.x users is a non-goal for people like regebro
[23:07:06] <regebro> dash: Right, because improving the experience of Python 2.x is what Python 3 did. So that already happened.
[23:07:31] <regebro> dash: Well, that's then not Python 2.8, which is what we are supposed to discuss here. :-)
[23:07:43] <richmoore2> dash: i get that impression too. and i can sympathise - as one of the qt5 maintainers it's annoying to get lots of qt4 bugs. but the fact is most users use qt4 and python2 so i have to live with it
[23:12:06] <richmoore2> if someone wants to have a serious discussion about how to make a better python 2 (particularly improving the libraries) then gimme a shout. for now, i'm going to step away from the computer
[23:12:11] <regebro> And should things change to it *is* likely to happen, what should be in it is pretty clear, and doesn't really need much discussion: Everything that can be backported compatibly.
[23:12:13] <dash> Anyway I believe pypy folks have expressed willingness to accept feature patches for their 2.x implementation
[23:13:42] <dash> regebro: There's nothing that python 3 more improveable than python 2, other than people believing it to be true
[23:14:04] <regebro> richmoore2: How many times do we need to say this: Until Python 3 is declared dead, the only way to add features to Python 2 is to gfork it.
[23:14:57] <regebro> dash: That the core maintainers are not interested in adding features to Python 2.x
[23:15:54] <regebro> Since they are not interested in adding features to Python 2.x, it won't happen, unless a whole bunch of other core maintainers show up that are interested and can show themselves capable.
[23:17:10] <dash> well, i'm ready to use a language implementation that isn't stuck in the 80s anyway
[23:18:23] <regebro> dash: For all your complaints about Python, and explaining how much Python sucks and that nobody is doing anything about the fundamental desig issues etc, I really have no idea why you actually are here, unless you are here only because you like to complain.
[23:19:13] <dstufft> regebro: probably because he uses python2 a lot, and he'd like to see a python 2.8 that solves his problems :)
[23:19:14] <regebro> "Discussing potential for a Python 2.8. What could it contain? What would it be like? How could it help moving people to Python 3? "
[23:19:53] <regebro> dstufft: But if you change the fundamental design issues in Python 2, it won't be compatible.
[23:20:21] <regebro> richmoore2: You are trying to discuss how to not move the Python 3. I don't see that in the topic.
[23:20:47] <richmoore2> potential for a python 2.8. what could could it contain? what would it be like?
[23:21:02] <regebro> No, just getting annoyed at having to repeat myslef infinitely for the same 2.3 persons.
[23:21:39] <richmoore2> has it occurred to you that there's the possibility that they might disagree with you and be correct (!!!both at the same time???!!!)
[23:23:10] <richmoore2> i'm not going to argue with you. if you want a civil conversation about it then try asking tomorrow (after removing the chip from your shoulder)
[23:24:00] <regebro> I'm just annoyed of having to repeat myself, and that you and others are determined to not be constructive.
[23:25:02] <regebro> eriknw: "the goal"? His goal seems to be just that. It's also quite definitely the goal of dash.
[23:25:42] <regebro> If we want the benefit of Python 3, the discussion should be how ww can make that easier.
[23:26:09] <regebro> This is what I'm trying to discuss, instead of having a meta-meta-meta discussion about it.
[23:28:08] <eriknw> what is "that"? some of the benefits of python3, or python3 itself? nevermind, I'm out