text stringlengths 0 2.35k |
|---|
• Engineers have different preferences for their work: some enjoy creating new code, while others prefer maintaining and improving existing systems. |
• There is a spectrum of engineering types, including maintenance engineers, prototype engineers, and those who do a mix of both. |
• Legacy system maintenance is often seen as a necessary but underappreciated task that requires balance in an organization's priorities. |
• Advocating for legacy system maintenance requires framing it as a financial investment to prevent future technical debt and improve overall productivity. |
• Quantifying the costs and benefits of legacy system maintenance can help teams get buy-in from product, business, and other stakeholders. |
• Differentiating between technical debt and legacy code |
• Quantifying and prioritizing technical debt for business decision-making |
• The need for new roles or positions focused on technical debt management |
• Using probabilistic language to communicate technical debt risks to non-technical stakeholders |
• Critique of current Agile methodologies and their inability to effectively manage technical debt |
• Preventing legacy code through testing |
• Importance of clear intent and documentation in code |
• The value of taking time to think critically about code and design |
• Pair programming and group design as potential tools for preventing legacy code |
• Investing in infrastructure and monitoring before rewriting code |
• Emphasis on small, incremental changes rather than large-scale overhauls |
• Legacy codebases and the challenges of rewriting them |
• Importance of understanding why a rewrite is necessary before starting one |
• Considerations for avoiding legacy code in future projects, such as designing with modularity and loose coupling in mind |
• The risks and complexities involved in rewriting a large-scale system |
• Cases where it may be more practical to fix specific problems within an existing system rather than rewritting it entirely |
• The challenges of rewriting legacy codebases and the potential for failure |
• The importance of having a plan before embarking on a greenfield rewrite |
• Kris Brandow's analogy of "mud pit development" instead of "greenfield" |
• The need for maintenance and upkeep, rather than relying solely on greenfield rewrites |
• Unpopular opinions and perspectives on various topics, including the ethics and philosophy behind technology |
• Jeff Hernandez's dislike of yogurt, including Greek yogurt |
• Kris Brandow's preference for Greek yogurt despite initially disliking it |
• Misha Avrekh's claim that CSS will someday replace all other programming languages |
• Jon Sabados' opinion that adding generics to Go was a mistake |
• Overuse of generics can lead to bloated code and maintenance issues |
• Balance will be restored eventually, but for now there's a lot of experimentation with generics |
• Abuse of concurrency features (e.g. channels, goroutines) is likely to follow, similar to the early days of Go |
• The van Neumann architecture is outdated and holds back hardware capabilities |
• C-like languages, including Go, are not well-suited for modern concurrent programming |
• New language architectures should prioritize parallelism and concurrency, like Erlang |
**Angelica Hill:** Welcome to Go Time! Today we're gonna be talking about how to prevent legacy code from creeping in. I'm joined by a wonderful array of guests, as well as my lovely co-host, Kris. Hello, Kris. |
**Kris Brandow:** Hello! And yes, that is my voice you hear out there, listeners. I am finally back, after four months. It's been quite a while, but I am very excited to be back. |
**Angelica Hill:** It's been far too long. And along for the ride with us, talking about how to work with, avoid issues arising from, as well as just how to deal with legacy code on a day-to-day basis, we have firstly Dominic St-Pierre, who operates a small consulting company in Montreal. He is the maintainer of a FOSS... |
**Dominic St-Pierre:** Thank you. |
**Angelica Hill:** Pleasure to have you. |
**Dominic St-Pierre:** Happy to be here. |
**Angelica Hill:** Then we've got Misha, who is a first-time Go Time guest, but also (I'm told) a first-time podcaster in general. Very privileged that you decided to grace us with your first podcast. How are you today, Misha? |
**Misha Avrekh:** I'm good, thank you. |
**Angelica Hill:** \[03:57\] And then we have Jeff, who is an associate backend engineer at the New York Times. I should also say, Misha is also at the New York Times and is a software engineer. Welcome, Jeff. How are you? |
**Jeff Hernandez:** Great. I'm excited to be here. |
**Angelica Hill:** Excited to have you. And then Jon, we've been here before... Jon Sabados, who is also at the New York Times. I promise I'm not biased. They just happen to be people I spend most of my life with... He is staff software engineer. Welcome back. How are you today? |
**Jon Sabados:** Doing well, and good to be back. |
**Angelica Hill:** Awesome! So before we dive into the nitty-gritty, I wanted to ask the group, what is legacy code? How do we even define what we call legacy? Dominic, do you wanna dive in? I know you've been working with something that I'm sure we would all agree is legacy... |
**Dominic St-Pierre:** Yeah, I think in my case it's really a "passion." To me, legacy is when a company starts to have difficulties hiring developers to work on their software, you can think of that as legacy, I think. Also, when there's no unit tests, not really good documentation, it's very old software that nobody ... |
So yeah, to me when the tooling starts to not work for you anymore, and things like that. In a nutshell, that's a little bit my definition of what legacy software is. |
**Angelica Hill:** So you said very, very old... So we're on Go Time. Go to me doesn't feel like a very old language. So I wonder, maybe - Jon, I know that you've had a little bit of experience working with this... But at what point did Go applications turn into legacy? |
**Jon Sabados:** Well, that's a good question, because you can actually -- and again, it depends on how you define legacy, but at least in our scenario it's very possible to have fairly recent code that is suddenly legacy. One example of where you might hit that is if you have a system that was implemented in a fashion... |
**Angelica Hill:** Okay, so any code that I don't wanna work on, I can say "Oh, it's legacy. It was written last week, but it's too old; I don't wanna work on it." \[laughs\] Okay. And then I wonder, Jeff, specifically - I know you've joined the New York Times preferably recently, and I know that you came to a kind of ... |
**Jeff Hernandez:** I feel like part of it is everyone just says that this portion of our services are really difficult to work with, they're kind of considered legacy. But I don't know, I feel like maybe my own definition of what is legacy is more broad than what's been given so far. Legacy is definitely for me someth... |
**Angelica Hill:** \[07:59\] Yeah, for sure. And then obviously, when we're talking about legacy, I feel like not only are we talking about legacy architecture, legacy infrastructure, but we have legacy dependencies. A lot of our apps have hundreds possibly of external dependencies. At what point and how do we go about... |
**Misha Avrekh:** Right. Well, for example the project that I work with now, which is written in Go - it's about two years old at this point, so probably not legacy strictly speaking, although there are parts of it that are starting to feel like legacy code... I can talk about that separately later. But yeah, as far as... |
**Angelica Hill:** And is that something that you do for all of your dependencies? Or is that something you kind of have a specific list, or -- like, how do you make that decision? |
**Misha Avrekh:** So on our team we made those decisions specifically based on our latency metrics. It doesn't really matter whether the external dependency was written in Go, or Java, or what have you. If we feel like we need to optimize for latency, we try to cache it. That's been our approach. So yeah, monitoring an... |
**Angelica Hill:** For sure. And then, obviously, when we're talking about legacy, in its many different definitions, because as we can tell from even these initial definitions, there is kind of a fluid way of defining, we can't just identify and then say "Great! We're gonna rebuild, refactor..." We have to maintain it... |
I know Kris, our wonderful co-host, has been kind of spearheading a maintenance series, so I know this really tees into the various different topics you've been exploring... Perhaps for our listeners who aren't fully familiar, I'd love to kind of pass over to you before we dive in, to talk a little bit about how we've ... |
**Kris Brandow:** Yeah. I guess I would start with saying my view of maintenance is definitely not in some ways as maybe negative as everybody else seems to think it is. When I hear the term "legacy", I think there is this feeling of dread that a lot of us have, but the term legacy can actually be a good thing, too. Li... |
So I think before it gets to production, when you can still change it and mold it and do what you like with it, I think at that point - yeah, we're still dealing with something that's not legacy, it's not debt yet, it's not something that we have to necessarily deal with in the future... But once it hits production, on... |
Now, of course, we can get into the semantics of what does it mean when something actually gets to production, but I think that's the way that I at least conceptualize this... Pulling it further back than anybody else has, and really seeing it as all of our code is legacy, and it really comes down to which code do we t... |
**Dominic St-Pierre:** \[12:01\] Well, yes, I really like what I'm hearing so far... One thing that jumped to my mind is to me a Go codebase, even though it might not have been designed the proper way or whatnot, it's still easier to refactor, it's still easier to change the status of the codebase... Because there's pr... |
**Kris Brandow:** Yeah, and I think that's definitionally just bad code, at the end of the day, if you can't maintain it in any way... But I'll also say, I've run into lots of circumstances where we have lots of tests and they're not good. They give us a false sense of confidence. Or you have lots of tests and they're ... |
So I think even in your example there, it's like - yeah, tests definitely can make a codebase better, and generally they show the mark of a better codebase, but I think there's ways in which even those types of codebases can be a pain to work with. But I would absolutely say, a codebase that has zero tests, no document... |
**Dominic St-Pierre:** How do you know then if your test suite is going to be good for the long run? |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.