PyCon India 2025 Keynote: Fellowship of the Stack: Scientific Discovery with Python - Dawn Wages

Yeah, folks. So, let me introduce our last uh keynote speaker, Don Wedges. Um, I'll give a small introduction. Uh, Don is a dedicated uh member of Python community community currently serving as a chair of Python software foundation board of directors and a member of organization for ethical source advisory board. She is the director of community and develop developer relationship at Anaconda, a leading provider of data scient science tools and opensource distribution. Dawn is a vocal champion for open-source software and a passionate advocate for building welcoming and inclusive space for developers and contributors. She brings her experience supporting Python developers tools at Microsoft and a focus on nurturing the longstanding community that supports Anaconda's open-source core to the conference. Let's welcome her with a huge round of applause. Hey everyone, I'm a bit short, so I'm going to be spending part-time behind the podium and a little bit of time out here. Also, it's the end of the day, and I appreciate seeing so many faces in the audience. This is going to be a little bit interactive. I'm going to ask for you to raise your hand um and kind of do a back and forth with me. So, if we could keep the energy up for the last part of the day, I would appreciate it. that would make the keynote go great. So, as we mentioned, my name is Don Wages, she her pronouns. I came all the way from Philadelphia to Pyon, India, and it's been a fantastic occasion already, and this is my first time in India. So, I really appreciate the warm welcome I've had. >> [applause] >> So today we'll be talking about the fellowship of the stack uh a scientic scientific discovery with Python and this is going to be a talk that's going to be really accessible to almost all of the audience regardless of your expertise. Um, but this talk actually started out and I want to give a shout out to to Pur Purdue uh school of advanced computing uh for letting me do this talk in a different iteration um before coming here. It actually had no narrative arc at all. And this is almost entirely about characters. Uh an arc on how an epic journey is going to get you to your in destination of scientific discovery with Python and how we can do a better together story. So my name is Don Wages. I am the outgoing chair of the Python Software Foundation. If you have questions at the end, um hopefully I go with enough time so you can give questions. Please feel free to ask you questions about the PSF. Uh we have ongoing elections going on right now. Um I will ask you to get your phone out if you can uh because there will be two QR codes that I would love for you to action. Um I also uh am a Python engineer and I have been for uh going on a little bit over a decade. Um, I also organize Python events and I've been doing it as long as I've been learning Python. Um, and I also uh contribute to the Wagtail CMS core team and the other things we talked about. I lead a team at Anaconda uh for developer and community relations. So, if you could all take out your phones, I Oh, I see you doing it. Yes, great audience. Yeah. So there are two CTAs main CTAs in this event. The first of which is I would love for everyone in this audience to become a Python Software Foundation member. There are two different ways to become a Python Software Foundation member. One is you can be a contributing member which means that you give five hours on average a month or more contributing to an open source project or contributing to an open source community in Python. If you do that, it is self uh nomination. You get to attest that yourself. No one's really going to like follow up on you to make sure you're doing it. It is honor system. But please give back to the community and you can also be a Python Software Foundation member. And then the second way is you can also uh be a sustaining member which means there is a sliding scale membership and you can pay to be a member of the Python Software Foundation with as little as $25 US. The other CTA is the taking the Python packaging survey. Uh this is a labor of love of mine. It's been years and we've had trouble getting off the ground a Python packaging survey. There's the JetBrain survey that is general Python ecosystem. This one's specifically on Python packaging. If you've ever pip installed, if you've ever installed, if you've ever uh cloned and and built a Docker container and with an automation script and your Python packages were already in there and you didn't have to touch it, you're still welcome to take this survey. We'd love to know how people are using Python and how they're interacting with their PI um packages in the ecosystem. We currently have 16,000 respondents, which is really big deal. The survey closes at the end of December and then there's going to be an open data set in early 2026 with all of the responses available. So your responses could be included in that data set and it takes usually on average like 10 to 15 minutes to complete. Uh, perfect. All right. So this is going to be our epic journey. We're going to start with why the scientific Python community uh and ecosystem is so treacherous to maintain. It is something that's widely known across the ecosystem that scientific Python and Python in particular has environments that are kind of hairy and have some rough edges. When it works, it works. Um but when it doesn't, it's really hard to navigate how to rectify and sometimes you just want to uh blow up the environment and start over. Why is that? um this discovery the the presentation through our scientific discovery it always starts with a spark uh in this journey we'll start to understand why the Python ecosystem is so treacherous and I posit today that it's one uh that there's two reasons for this one there's the evolving nature of software um and it happens across software not just in Python and we're in a critical inflection in in the software development ecosystem because there's an incredible investment in making the development process easier for us. This is from the soaring success of VS Code. I mean, we probably have some Emacs and some Vim users out here. Um, but IDEES are really the centerfold of how people are interacting with their code these days. And also the more recent advancement of AI assistance making it even more efficient for us to engage in what we do. And the second uh is because you can do so much with Python. So in the early 2000s with the proliferation of Python across many different uh subject matter expertise, you saw because of the way that Python provides you a way to interact and build with a level of abstraction away from the machine, you're able to build faster and more performantly uh performant in a different way. It's not the fastest code language, but it is more performant to build something quickly. Um and we're empowering users to build faster. That also means that there is different and disperate directions that the community can to uh can go to build one thing and it makes managing those uh competing dependencies and environment sometimes more difficult. So it provides an immense amount of good and also allows for a diverse group of developers to create their very specific scenarios. We also want to go into uh oh let's has I would love a raise of hands. This is first of interaction part. Who's heard of the Python documentary? Oh, yay. Good. Who's watched the Python documentary? Okay, good number of you. So, we're going to do a little clip of the Python documentary right now about how the first Python uh preversion 1.0 was distributed. If I the release which turned out to be an incredible pain because Usenet was the only thing we had to release it. [music] Usenet was a network of you could say bulletin boards. People would write messages. They would get distributed over the world and then there were the groups for [music] all the different computer languages. Usenit had a very strict limit on the size of a single post. First [music] put your entire source tree in a tar bottle. Then compress it. Then you have to do another step where [music] you turn the binary data into ASI encoding. So you have this huge file with gobblygook, but it's at least printable gobly gobblygook. And then you apply another tool that snips [music] it into pieces. Then there is a script where if [music] you have downloaded all 21 parts, you can undo all those operations in reverse. That was what people put up with. So we did all the work on our side and [music] plenty of people did all the work on their side to unpack this because I had apparently written a very good teaser about what this was. And that's a little bit of the um let's put this back into presentation mode. Perfect. All right, we're back. Okay, we're back. Perfect. So, um that's a little excerpt of the documentary which I really enjoyed watching. Hopefully that's enough to get the rest of you to want to watch the rest of the documentary, too. And it's all on YouTube. Also, we're going to dive into what the ecosystem is doing currently to address it. So, from uh really difficult developer experiences like we just saw all the way to what we're going through now, we're iterating on a better developer experience every day. Um but you know there's much more to do and then where do you fit in a scientific Python fellow um how can you advance this? So in the grand scheme of this uh metaphor we're going to get our supplies. We're going to cast our first spells. We're going to gather companions and address some of the trials and tribulations of the Python ecosystem. uh we'll form a fellowship along the way and then we'll scale our quest. We'll give you some extra activities that you're going to do to continue this advancement and take record of our exploration. So you'll need a language distribution. There's many different ways that you can get your Python. Most people are more familiar with uh CPython. That is the uh default distribution that's done by the uh Python steering council. There's a language interpreter manage package manager development environment. You'll also want your version control. Um, and if you're doing uh exploration with data, you'll want your data input and your pipeline for ETL. And then your specific domain is going to determine how your exploratory toolkit or your analysis engine is going to work. So we uh stumble upon a guide and we a shopkeeper of sorts um and in their own right is experienced with all these very and difficult tools to help us on our journey. Your IDE is your interactive development environment will transport you to another environment and keep track of your it'll keep track of your variables um of that universe and connect symbols across your files um into other portals. Uh I recommend a health check of sorts. Tools to provide uh health checks on your code are invaluable. For example, cond forge and mamba have environment health checks. You can also consider hypothesis testing, data alerts on your pipeline uh to identify data drift um or data ch data quality track uh quality experiments um and keep tracking of those. And this may look like a simple ball of twine uh but it is uh what holds your code together uh from your developer environment to your staging and production scenarios. Twine the tool that um that is under GitHub under the hood uh keeps your CI/CD moving from one environment to the other. And don't forget code for matters llinters and style checkers. uh organize your imports in security focus on linting uh in security focused linting like bandit. Your spell book or your documentation starts with writing spells so that you can recite them truthfully every time you need them um to sit down and embark upon your own journey. But soon you'll expand that spell book to uh be helpful for others. Share it freely and reduce the barriers for your fellow journeymen. uh and people across all levels will have the same reach that you do. And think of this like a time machine of sorts. You have mysticism, but beyond what you know, specifically in scientific computing and scripting in Python, it's easy to jump into code without justifying why you should include version control. Think of it like your time machine. And the magic of data. Don't forget your data. your data is gold. So, when we're casting our first spells, another interactive portion, I'm going to ask you all to raise your hand. Um, as you're starting your projects, are you who who in this audience is using languages other than Python in their Python projects? Okay, I see a number of hands up. Um, are you able to or do you typically anticipate the dependency considerations before you start? Um, or are you uh stumbling upon the dependencies and solving it as you get to it? So, who's considering it before they even start the dependencies? Who stumbles upon the dependency uh contradictions and then fixes it as they go? Okay, more hands there. What about sharing environments with your peers? Are you predicting or expecting to share environments with your peers hands? Okay, a good number of people are are doing that. Um, and what about tools that use projectbased approaches like Pippen, Pixie, UV, poetry? Hands. Okay, good number of people doing that as well. Cool. So this is like a common journey that you'll have in your data science and scientific computing um process. You'll want to start with identifying your problem and your scope and then get deeper into your data understanding. Many times cleaning and uh normalizing and wrangling your data. You'll model your model your data when you uh have data science variations in scientific computing. Um and then you'll evaluate um that uh those models and then check back with your production results. And here we we met a friend with deep knowledge driven by many years of p pursuit wise in many years seemingly magnanimous. Uh but it's their desire to understand the universe that makes them a great companion. And they're well-versed in the new and the old magic. They hear about your journey and they're immediately interested in helping. There's so many different ways people use Python and there's a huge ecosystem with hundreds of thousands of packages available in open source to exped expedite your innovation. So you're able to leverage these packages for your accomplishments, speeding up the time to finish your task and pushing away from your desk. Uh the first generation focused on integration with Python uh Python's ease of use with the performance of compiled languages. Roughly from the mid 1990s to the early 2000s uh tools like F2PI which would be for Python swig and later CPython um really fit that brill uh there was an emphasis on arrayentric computing introducing in dimensional arrays that addressed uh Python's native list limitations and numerical data enabled mathematical operations uh that are on entire arrays without explicit loops. So these free tools that were alternative to expensive pri proprietary tools um like mat lab for example and this is all within the academic and research origins. These are developed by scientists for scientists. Um and that signaled um a a big shift in the Python ecosystem. And by the end of this first generation uh we we see a movement into some of this new magic. The most newly adopted and widely appreciated packages in 2024 2025, excuse me. Um Polers is one of the rising stars that hit version 1.0 0 and is rapidly gaining adoption in potential pandas replacement. It's lightning fast uh dataf frame library that offers significant performance improvements. Uh performance particularly like polars and ductb are in a league of their own um being in an order of magnitude faster than dask and pispark. Uh there are streaming engines that are can seem to be three to seven times faster in polar's in-memory engine. And then there's also a huge wave with GPU accelerations. So for example, I've been playing with the NVIDIA Rapids library which has been really exciting and that uh accelerates, excuse me, accesses your uh GPU whether remote or cloud um and provides up to 13 times performance improvements inside the Rapids accelerator polar's GPU engine. Um another uh approach has also been uh with ecosystem integration. So for example, polars uh works with major packages including scikitlearn and plotting libraries uh through narwhal's compatibility layer. There's also high performance data manipulation in Python. Um so all of these new features have led to this new new magic of sort um that's accelerating where we're going. We're seeing some changes with a performance first approach. Uh there's also a zero code change philosophy. The most successful new packages of 2025 focus on providing performance improvements with minimal code changes. That would include the Rapids library. That's really exciting. Uh AutoML and democratization. So automated machine learning tools are becoming more sophisticated. Um and they're making advanced analytics access to be uh to non-experts. So, we talked about our scientific companion. We have a new companion that we've met ready for this adventure, showing their fortitude. They're bringing in software competencies. This includes uh numerical accuracy, testing, uh being thoughtful on resource optimization, creating dry code. For those who may not be familiar with dry, that's when you don't repeat yourself in your code. So that's levels of abstraction and uh uh areas of um ownership and expertise of your code uh division of responsibilities of your code also platform independent. So it can it run on one machine versus another maintainability which includes code quality dependency management your CI/CD pip pipeline and your documentation. Some of the key insights is that uh scientific computing serves these dual purposes. It must produce produce correct scientific results and it must be maintainable as a software system. So this dual requirement makes these principles especially important in the scientific computing ecosystem. Um I do have a call out at the bottom. Who of this audience I'd love a raise of hands who's familiar with PI open sigh? Okay great a few of y'all. So hopefully I am introducing even more. So if you are of uh in a scientific computing background and may or may not have experience with Python packaging for example um this is an organization led by Leo Waser um and they're really fantastic group that have tons of trainings and they bring in experts all over the ecosystem to help you with your Python packaging specifically in a scientific computing context. So let's dive into some of the trials and tribulations that we may find in our scary ecosystem that is that can be Python. Python support is a lot to learn especially in scientific computing when your first domain may not be computing. Python's considered a glue language. So for example um there's a lot of people who call Python the second best language for everything. uh that would mean that maybe it's not the fastest language. Um but because it can do a lot of different things, it's very flexible and it is a a great middle layer language between your many languages. Um it is definitely a a popular language to glue your two languages together. Maybe you'll have a Java backend and a JavaScript frontend and you're running pipelines or integrations with your pipeline. It's used in a lot of different use cases. It's very flexible and extensible and it can outsource a lot of the performance bottlenecks. We'll get into uh the next conversation we'll get into the Python interpreter and we'll talk about how the C API uh works in Python. And also it's human readable code. So I think that's a large part of even why the documentary but a long time why people say that Python has gotten so much adoption is because you can write your code walk away and not look at it for a year come back to it and still decipher what you were doing. So we have a side quest and we have a special friend to help us through our side quest. We're learning the intric intricacies of the CPython interpreter. who would consider themselves among this audience pretty familiar with the how the CPython interpreter works? Raise a hand. Awesome. Great. So, it sounds like there's a good number of people who are coming along for the ride and some people who may learn something from uh from this little side quest. So, we'll uh there are many Python interpreters. There's Jython which is written for Java and uh written in Java for Python. There's Iron Python which is written for.net. Uh there's pi which is for just in time compiling of python. Uh it directly executes the code rather than needing a a separate compilation layer. So that so that is why um python um is directly as as this. Um so either through hitting a play button in your editor or shift enter or uh python main on your CLI. There may be other ways that you can execute your Python. We have this friend whom we met along the way who can help us break down this process. Uh Python is an interpreted language. So not all languages are considered interpreted. Uh some are compiled and that means the program that you run is delivered as source code and you can just run the source code. It'll start out uh with your Python file uh by the Python interpreter without any fiscal physical compilation. This is human readable code. This is peri Python performing the lexical evaluations and then breaking it down into character tokens along with keywords, identifiers and operation operators and literals. The tokens are now representing the shape of the code. Then it means uh excuse me and then you go into your Python interpreter. The interpreter takes and executes each statement written in your source code. This will be line by line. uh the program gets compiled by the Python compiler um and then checks for errors and if the compiler finds an error it will show the error in the console and if there's no error uh the source code is well formatted then they'll be converted into byte code. Byte code is the low-level unbiased representation of your source code. So the compilation to the bite code is accomplished through the Python compiler and the bite code the Python bite code compiler. The bite uh code is then processed inside the Python virtual machine and it is interpreted to get give actual machine code. So in order for this portability of the source code to be valuable, this Python virtual machine now allows us to bring uh allows us things that happen at runtime like dynamic typing and runtime introspection. Uh this is also where the memory manager and the garbage collection lives. Uh PEP 744 proposes a departure from uh the PEP 659 which is specializing with the adaptive interpreter and this rewrites the bite code uh instructions in place. So we're able to have more visibility into the profiling of the code. For example, we would be able to see what paths are being executed the most or the memory layout of particular objects. And this is where the C API happens. So if we're offloading highly uh bottlenecks in our code, we're able to leverage the C API. This is dynamic loading of li libraries um in the standard OS method. So it'll load the entry point in its naming convention and then it invokes a functionality where it's entirely responsible for what the modules set up and own. This is important when I talk more about optimizing with spy or numba. There'll be a little bit more about that later on this presentation. Um but this is a a nice extension point for when you have C, C++, forran which makes it relatively easy to lean into performance improvements. And then you have your machine readable code. The interpreter never produces any intermediate machine code unlike a compiled programming language. And Python is easy for development environment because of the way it that it shares the source code and interpret it dynamically. It'll allow you to write your code in all sorts of ways have all different types of tedious and difficult dependencies and Python's interpreter will just try its best to handle it. And then there are packaging complexities. We meet another friend who's going to help us with another tedious uh part of our main quest. Uh this is these are deeply nuanced and scary complexities that offer and they decide to offer their help in understanding in Python packaging. combining a utility into a reproducible and distributed version control way that allows for iterative improvements uh can work uh can be work fine or it could be incredibly difficult. Sometimes a package can be buil built differently based on the date or maybe bugs in GCC which leads to loss of numerical precision based on import order or the having the correct configuration sphsl in particular their configuration script is in pearl or maybe shared symbols across libraries that can delete reference differences between versions silently. In GDAL requirements, this requires building Chromium, for example. Sometimes it'll build differently based on file system issues or it'll build differently across architectures. These are nontrivial considerations. For example, pillow relies on an external C libraries like lib JPEG, lib png and zlib in free type. So that adds additional complexities to packaging pillow a commonly known kind of headache in the ecosystem. And this is not necessarily to denigrate any of these packages. Uh these are how legacy concerns happen and these are known concerns in the ecosystem and how to address them are not easy. in GDAL the geospatial data abstraction library which is also notoriously different difficult um it has complex geospatial toolkit with numerous dependencies and platform specific quirks so for example GDAL supports dozens of geospatial formats which all require different libraries uh PJ coordinates transformations geos which is geometry operations there's SQL light postgress HD HDF5, net CDF and others. And it also imports uh different geospatial formats like geotiff, shape files, postgis, etc. And pietorch uh these conversations are happening internally and out in the open uh for the community part to participate. For example, matching pietorch versions with your CUDA versions. PyTorch skips ships with specific CUDA versions baked in and there's a huge download size uh with PyTorch and there's also platform specific wheels. PyTorch has strict compatibilities with requirements of NumPy and other core libraries which conflict with other packages in your environment. And then we meet new friends, ones with domain expertise are imaginical siblings which require you to make decisions before you can deploy their services. You go down into domain expertise or does your general purpose and build your own suit your needs. uh we'll we could start with a general purpose uh tools like numpy and sci and then you can narrow in deeper into the funnel like broad scientific uh packages like pandas or mattplot lib uh we have domain specific like biopython astropy scikit image and then you'll also have maybe lab specific uh packages that meet your specific needs um in custom analysis these are all tend to be problem driven and designed and aligned to your data types and formats and incorporate domain nuance that allow for interdisciplinary uh collaboration. So our fellowship is forged. It's 2025 and scientific computing the underpinning of AI is finding its most salient applied use cases in this space. We have language and package distribution, your IDE choice, your framework architecture, um environment standards and artifacts, and your build process and maintenance. The creators and the most active consumers have forged allegiances over the last 30 odd years. And some of these trends have emerged. So when it works, you may not notice when your Python distribution is not CPython, but there is a proliferation of other Python distributions. I'm most familiar with CPython and the Anaconda distribution and PIIE, the just in time uh Python distribution, but there's about a dozen more dependent on your use case. People like development environments. We've talked about this a little bit before. It's gotten deeper in your expertise. Um and then people are also familiar and interested in tools that or orchestrate your workflow. These are seen across the ecosystem and there are too many environment standards in artifacts and a huge proliferation of tooling. I wanted [clears throat] to highlight three different PEPs for people who are not familiar. PEP is a Python enhancement proposal. Um so PEP 440 uh that's about version identification. Um and if you're interested you can go deeper and to look into some of these PEPs. think it's really interesting to um to follow the PEPs and where they go and the conversations around them because you'll get deep expertise on why in the architectural decisions um made by the uh proposers. There's also PEP 621 which is about product met uh project metadata and then there's PEP 518 which is build system requirements. All of these are more recent or ongoing PEPs that are in um discussion that people are really interested in to advance our Python ecosystem. And so we met our helpers. We had we started our journey. We met a friend who showed us the way and inspired us to believe what was possible. An epic journey. We met a shopkeeper who showed us the tools and a sage friend with deep scientific knowledge of the old and new ways. Uh we had a formidable warrior who equipped us with the software engineering pre principles of resilient code and two deeply technical companions. One who helped us navigate the Python interpreter and the other who helped us dodge the perils of complex packaging dependencies. And last but not least are f friends with deep domain knowledge. They know the dark arts of every form and can help us navigate our own domain nuance. But where do we scale? So the next step in uh your packaging journey is going to be um making sure that you have observer observable, monitored and logged code. Uh you'll also want to have enhanced collaboration tools. So this would mean um having uh reproducible research environments of sorts. Um sometimes you can have tools that will allow you to work on code and jam together. Um, and then also some tools that will create hashes or artifacts uh that will tell you if your code environment have changed. You'll also be interested maybe in running on specialized architecture. So if you have high performance computing needs um and you have um compute intensive jobs, you might want to put that on a compute cluster somewhere on someone else's code uh excuse me someone else's computer or you might want to be have uh use your GPU and have CUDA enabled computations. I'm also really interested in the productize before production movement. Um this is the uh build it so it works for your specific need and then do another round of refactoring to abstract away some of these dependencies, reduce repetition and um maybe um pull away some code that can be uh open sourced and used and useful for others. I'm also interested in module workflows, modular workflows, uh ones that can be uh cut up and you can document them individually and hopefully when done well you can find drop in replacements as you uh evolve. And then the best workflows scaling your quest is always well doumented uh workflows. So that was scaling your request with code. I want you to scale your quest with the PSF. So there is an opportunity to vote in this upcoming Python Software Foundation election. You can also join as a Python Software Foundation member. And then I'll talk a little bit more about the funding cycles in the PSF right now. So if you're not already involved or have not gotten your email, if you are a Python Software Foundation member already, um you should have that email for OPA vote. closing uh voting ends uh at on September 16th, 2M UTC. But if you're not a Python Software Foundation member and if you're just interested in seeing how the process happens from the beginning to the end for a full calendar year, you're in a great spot. You can sign up to be a PSF member today. Here's a QR code. This is where I got you originally to take out your phones and you can be as PSF member. you can see the entire election cycle and then you can follow along to see the uh board release minutes and um and voting procedures and learn a little bit more about the PSF before you get a chance to vote in the 20207 elections. And last but not least, uh I wanted to talk about the funding in the Python uh community. So, Python is known, the PSF is known for giving hundreds of thousands of dollars every year to various communities across the world to help them with their Python events. Um, it's something that we pride ourselves on. It's a main tenant of what the PSF does to help proliferate the the language across the world. Um, and we've paused our PSF program. There is a screen grab of the register talking about how the growth of the Python language um has not uh led to the same growth in investments in the foundation. And so the pause has been twofold. It's been because we saw in the tea leaves we've been paying attention and we knew that the demand and the request from the ecosystem on us for more support has continued to grow. Um but we also saw that we've gotten into a more expensive conference environment. Um our the main driver of funds for uh the PSF which is PYON is was uh not returning what it used to. Um and then we need the reciprocal investments from large companies and we've seen similar movements across the ecosystem. All PI PYON conferences are many PYON conferences are experiencing the very same thing. So, we're taking a pause um for a full calendar year. It is unprecedented, but we're really excited to get back to funding the ecosystem in a new and more sustainable way. And in the same vein, we have an endofear fundraiser. So, please keep a uh your eyes out for pyfound.blogspot.com. That's where you'll find our announcement for the end of year fundraiser. And if you have a company match, for example, you can fund the Python Software Foundation or you can throw in the amount of a coffee every month um in order to help fund the organization that uh proliferates Python across the world. Great job. We did it together. We unlocked the scientific uh the secrets of scientific Python. >> [applause] >> There's a few ways this analog Oh, almost done. There's a few ways this analogy falls short. So, documentation exists. It's not dark magic. We iterate. It's not just one long journey. And you do not have to understand the entire ecosystem to be successful. Thank you so much to Risa Erllin uh who um uh provided the beautiful art that inspired this uh presentation. And here's another uh link to the Python packaging survey to take and you can follow me on socials. Now I am all done. Thank you Don for the amazing note. Thank you. >> Um if you want to keep it up on the screen, I'm open to questions. I had open questions at a previous event and I got grilled about the PSF funding cycles. I'm very happy to talk about the PSF. I'm happy to talk about Anaconda, developer relations, uh my journey in the ecosystem, scientific computing, any of any and all of the things. Any questions? I got one in the front. There's one right there if you want to or I can also repeat it for the for the audience. Awesome. >> What is the minimum age for becoming a PSF member? >> There is no minimum age for becoming a PSF member. >> Anyone can become. Great question. [laughter] Some like to know what the PSF does. Some wants to know what it's like being the chair of the PSF. Uh some have asked me why I'm not going to be the chair of the PSF in the next coming election. Why I didn't run again. >> Hello. Yeah. >> Yeah. Okay. Got one. >> Uh we pretty much talked about uh packaging. >> Yes. >> So pip was part of uh Python, right? Uh we we we get that as we install the Python. >> Yes. And now we have many things UV poetry. >> Are we having any proposal to include any of these tools into Python? >> No. So it wouldn't be to include any of the tools. So there is uh PEP 772 which is the Python packaging council. Um so originally the Python steering council which is the organization that manages all of the PEPs and that will decide what um exists outside of Python and what will bring brought into Python. There was one person who handled all of the packaging needs that handled PEPs. It was just not enough bandwidth. So what they did was they elected to have spread it out into I believe five people who will exist on the packaging council and they will decide um any packaging standards um right now what I know is on their docket um but they have a list of intricacies um is a standard packaging format. Yeah, back then we had uh easy initi then we have >> yeah I don't think they're going to bring bring in any of the existing tools. There is the PIPA which supports some of the tools like PIP and maybe poetry some others. Yeah. So yeah, I don't I don't believe they're going to be bringing in any of those tools into the Python ecosystem. >> Thank you. >> Yeah. >> Yeah. Hey. Uh, so I uh so I recently heard of this uh a new organization that formed called the Python Asia organization. >> Yeah. >> Uh is it separate from the PSF? What's the relation here? >> They're separate from the PSF. So the PSF is a North American nonp or it's a US nonprofit um that is uh its creed is to its mission to support Python everywhere. So it is within our boundaries to uh support externally but we are a US nonprofit and so we're governed by those rules and so there's uh Python Brazil as well and Euro Python the Europyon Society for example those are all uh considered sister organizations like Python Asia which is exciting yay Python Asia so that's good yeah not the PSF >> so how do the like how does the dynamic work like do they work with each other like is there a structure there or >> there's a lot of people who volunteer for the Python Software Foundation who also volunteer for Python Asia um but they are two separate groups and the interaction is um completely dependent upon the organizers and there's no formal interaction right now it's they we just leverage each other's resources and we're friends in the ecosystem and we share ideas >> that Okay. Yeah. >> Hey, good evening. Hi. >> First of all, a great session. Thank you for that. >> Uh, so my question is a little bit irrelevant to this. >> So it basically I am working in Golang as well as Python. >> Okay. >> So u we have seen a huge difference between how much Python can do and how much Golang can do in terms of concurrency. So the point is that uh are we doing something to improve that in the upcoming versions of Python or uh how are we going to do uh you know how do we optimize it uh you know to get better performance out of Python. >> Yeah. Yeah. So there's so I got good news and and bad news on that one. So there was a team that was spearheaded by Guido Van Rossim at Microsoft called the faster C Python team um that just came to a close this year. Um a lot of their performance improvements or some of their performance improvements will be um made it into 311 um 312 and 313. Um we'll also see some of those in 314 that includes uh garbage collection for example. I have a whole blog post on the other four that I think are really cool, but they're all very performance improvement heavy. Um, where the rest of the performance improvements are going to go, I'm we're not sure, but there's definitely PEPs um to facilitate this. People are excited about the JIT. Um, people are really excited about sub interpreters. Um, also free threaded Python. People are really excited about that. All provide some performance improvements. Uh just a followup. So uh I mean we are all here because we love Python. So uh any chance we should you know we should be worried about you know Python becoming a threatened sort of a language you know just like the old languages went away something like that. >> That's an interesting question. I am I'm of the mind that we should always remain vigilant on uh where um mind share could go outside of Python. Um I think there's a strategic advantage because I've invested so much of my time into Python. Um there is a strategic advantage to being careful and thoughtful and wanting to continue to make Python the best language there is. Um, but also I would love to challenge that question back a bit and my own um predition predisposition for wanting to keep Python great and say that whatever comes next, if there is something that comes next, a better iteration um is hopefully comes of the the vein of open source and innovation and um puts us in a place all of us in a place to pivot in a new direction and do more. So I I love open source. Python, just because I love Python doesn't mean I necessarily hate TypeScript um for example. And so if we do it in open source and if we do it in the spirit of collaboration, I'm really excited for whatever comes next. But I do love Python and I want to keep it great. So it's fair. Um I'm not particularly worried. I think that we absolutely have the mind share right now and it only looks like we're going to keep going up. Um, but I think the next wave is going to be making sure that Python is optimized for the browser. So, I'm super excited for Web Assembly being a P platform for Python. >> Great. Great. Thank you. Thank you so much. >> Any more? Oh, I see another one. Hi. >> Hi. Uh, I just have one. >> Okay. Uh, >> last question or we have to be done. Okay. I'm so sorry. Let's Let's go find each other in the hallway and I'm happy to chat. >> Okay. All right. Thanks everyone. >> Thank you, Don.