text stringlengths 0 2.35k |
|---|
**Kris Brandow:** Good design of your software. I think if you've gotta kind of push the boulder up the hill, or really go back to the genesis of the stream... Like, if you're gonna build a dam, it's a lot easier to build a dam across a stream that's two feet wide than it is to build one across one that's 200 or 2,000 ... |
But of course, a dam across two feet of stream is quite useless of a dam... So there's a middle ground here where we need to find the right place to put things. And I've definitely found that right place to be in good design documentation and good design ideology for the software that we build. But I'm sure other peopl... |
**Jon Sabados:** And I think I'd add a modern codebase does also not mean tested. There's plenty of fresh code that I see written that engineers oftentimes don't like to write tests, so that is a pervasive problem that seems to plague all codebases, is stuff just isn't tested. |
**Misha Avrekh:** Yeah, one thing I can say we've done to mitigate that problem is to keep an eye on the coverage thresholds and actually building the test coverage thresholds into our testing and deployment pipeline. So if someone adds new code without adding tests, the pipeline actually fails. |
**Jon Sabados:** That is definitely an interesting approach, but I've also seen it backfire horribly in the past. I've seen engineers, and this actually got someone talking to, this is in java land, put an entire class on a single line, because they were just annoyed by the test coverage requirements. So you can kind o... |
**Angelica Hill:** \[16:10\] I'm interested to hear a little bit more about -- you referenced Java, Go... Are there aspects of Go as a language that make it easier for you to either "avoid" legacy code, or avoid it being a whole huge refactor project? Because in fact, refactoring bit by bit, backwards-compatibility com... |
**Jon Sabados:** I think there's some, in that it's a simpler language. Up until recently you didn't have generics, so you're not gonna have people going nuts with generics and creating really complex things, like trying to deduplicate everything through generics... So that helps. But I think the biggest thing that Go ... |
That's also something we learned every time we write code and every time we develop, so there's a lot of patterns that we might have been doing 20 years ago in Java land, that we've maybe learned better ways than Go. But again, that's not so much a function of the language, just of the newness of it. |
**Jeff Hernandez:** I just wanna add that what Jon was saying kind of made me think about what this topic of legacy will be like for Go in 10-20 years, in terms of when it gets to be old enough to -- the same age as Java is currently, and we have a lot more Go projects around that need maintenance... It's just an inter... |
**Dominic St-Pierre:** Well, I can tell you about stories with .NET, even though -- let's say it started in 2001. So in 2005 there were already projects that you could not even migrate to the next version. So to me, the tooling that the language provides, which is very solid in Go, in my opinion, and the fact that it's... |
**Jon Sabados:** I guess I'd be interested, because I've never really worked in .NET. I've always been allergic to Windows. But the Java side of things, they strived for a long time to make the JVM backwards-compatible with prior versions of Java, and that didn't necessarily stop the need to have complete rewrites, or ... |
I've actually even seen the new thing with Go, where they're seeing projects that were running on old versions of GAE, that have pinned their version of Go to 1.11, or something, and upgrading those has become a nightmare. So again, it's less likely in Go, and it is for sure easier to generally upgrade versions of Go t... |
**Dominic St-Pierre:** \[20:09\] Yeah, but you are talking about the platform. So the Google App Engine is a platform; it's not really Go's fault or responsibility, in a sense. You can still take codebases from the early, early days of Go and compile that with the new compiler. This is something that you just cannot do... |
**Jon Sabados:** Yeah, and I guess there's a different past user story there, and experience, because - again, coming from Java land, everything will compile just fine, but the things that I've encountered, that I've seen, that make upgrading hard isn't so much the compilation, which it sounds like in .NET that might b... |
**Dominic St-Pierre:** Yeah, I can give another example. So you know, TLS 1.1 just vanished recently, but there is still old applications that run under an old version of Windows Server 2008, which is not supported anymore... But because their migration path to the next version of .NET is not an easy one - I'm not sayi... |
**Jon Sabados:** That does bring up is working in legacy with a closed source system is different than working in legacy with open source systems when it comes to those dependencies... Because that is one thing I've always liked about working in all the various different techs I've been in open source. You know if Orac... |
**Angelica Hill:** So do we feel like that's one of the differentiators of Go, in that we have this vibrant, wonderful community where if there's anything that is detrimental, it's causing people to feel that it's legacy in some way in terms of the tools provided? That there's gonna be mass opera in the comminuty sayin... |
**Jon Sabados:** It definitely can help, up until the point that there's enough bifurcation that there's a fork, which hopefully won't happen. It's rare to see languages truly fork like that. So I think largely yes, but that forking is also a potential danger. |
**Kris Brandow:** Languages except for JavaScript... \[laughter\] |
**Jon Sabados:** Oh, God, JavaScript... Yaay! |
**Break:** \[23:54\] |
**Angelica Hill:** So we talked about the pains of maintaining, and the need to - as you've kind of touched upon, Dominic, kinda of hack into these small solutions to make sure your apps continue running... But at what point do you throw your hands up and say "No, we need to completely rewrite this"? Well, one, rely... |
**Dominic St-Pierre:** One nice thing about legacy software is that at some point they kind of run without any intervention from any developer. So that is something that could help with a complete rewrite. If the company is able to do that and seize the value in doing that, then yes. I think at some point the maintenan... |
**Jon Sabados:** I don't know that there is a hard and fast rule for when to do a greenfield rewrite other than it should probably be one of your last resorts, because it is an incredibly expensive thing to do. Oftentimes you have to do it bug for bug, depending on your downstream clients, and that can be a nightmare t... |
And you know, different scenarios can offer different reasons. If you're doing a cloud migration - well, there may very well be a reason to do a greenfield rewrite of fairly recent stuff, in that case, because your underlying platform is changing dramatically. |
**Misha Avrekh:** \[27:41\] Based on the projects I've seen, there definitely needs to be a close alignment between business and product, as it were; business requirements and the tech requirements and what needs to be done on the tech side... Because that's when you start seeing possible solutions, like "Oh, maybe I d... |
**Jeff Hernandez:** I'm curious as to when it is important for business and product to be aligned, but also taking into account the engineering quality of life. If I'm spending all my time -- like Dominic said, I'm not fixing the code itself, I'm trying to keep the code standing, keep it running, because everything aro... |
**Dominic St-Pierre:** Yeah, I can totally relate with what you're saying. \[laughs\] |
**Misha Avrekh:** Yeah, if your engineers -- oh, sorry. |
**Dominic St-Pierre:** I'm alone at the company at the moment, so they are simply not able to hire anyone else. So I'm kind of stuck there. |
**Kris Brandow:** I do think there are some engineers who actually do like just tinkering on old codebases and not adding new features or anything, but just keeping something running. I think there are people that just get an immense amount of joy out of just doing that. But I definitely think trying to put engineers t... |
**Angelica Hill:** And do you feel like that is more in keeping with just like personal preference, or is that perhaps to do with where you are in your progression as a software engineer? I mean, purposefully we have kind of a range of different levels on this call, because I wanted to really get an understanding of wh... |
**Kris Brandow:** I think it's a different type of engineering, rather than a certain level of it. I think there are people that want to be maintenance engineers in the same way there are people that want to just be pure R&D, pure prototype engineers. They wanna build stuff, they never want any of that to be in product... |
I think there's like, okay, maybe you're on one far end of the spectrum, you really like getting one of those old, nasty codebases, and you're like "I'm not gonna add anything new, I'm just gonna fix it up. I'm gonna add tests, I'm gonna document things, and I'm gonna turn this into a really nice codebase to work withi... |
At the other end of the spectrum are those prototype people that are like "I just write code that I just clobber together and make something work. But there's no way this will work for the long-term." And I think the majority of engineers fall in the middle. They like doing some new stuff, they like maintaining some ol... |
I think that's one of the criticisms I have for a lot of tech companies right now, is that they are very much focused on the other end of that spectrum, that is the R&D engineers and the people that want to build newer things, and the results of that is a lot of codebases that don't wind up getting properly maintained ... |
\[32:19\] I think that happens, a lot of teams sitting there being like "Well, we really want to work on all of this old, legacy stuff, but where is the time to do it? How are we gonna make this happen?" So I think getting some of that balance is good. |
But to answer your original question - yes, I think they are different types of engineering, not different amounts of time that you've been doing software engineering. |
**Angelica Hill:** For sure. And off the back of that, people know that a certain platform or a certain part of that technology, that stack is legacy; it's known about. You're slacking about it, "Oh, this annoying platform. Oh, this annoying..." Like, it's all known. But how do you then advocate for a rewrite, or for t... |
**Jon Sabados:** Yeah, that's a hard problem, because effectively, that maintenance is a cost center, with no -- you don't see the benefit immediately. Generally, my experience has been everyone is short-sighted and more concerned about getting their current feature out than they are about the ability to add another fe... |
Most people understand the idea of debt; there's only so much you can have. And explaining it as tech debt and being like "But we will get to the point where all we can do is pay our tech debt, and we won't be able to make any new investments." And oftentimes, that falls on deaf ears, so... |
**Kris Brandow:** I definitely think structuring it as a conversation around debt and the financial aspect of it can help people, but also just like -- I think a lot of it has to do with process and project management, of like "Okay, when we said that we could deliver this feature, we said it would come with this amoun... |
I think it does take a lot of fighting and advocating, but I think part of the problem is that in general, the teams I've been on have been kind of not great at explaining exactly what that technical debt is. It's like, we took on some technical debt, but it's just kind of this mysterious thing. It's like, "Oh, you've ... |
So I think it's less that product and business don't want to pay for it, but the cost of it is not known, not quantifiable. It's like, if we talked in debt, but you never found out what the interest rates are. It's like, "Okay, well, you're telling me I have to pay down this debt, but is this debt that's at a 0.5% inte... |
Not being able to talk about things in that kind of quantifiable level I think is what holds back a lot of engineering organizations from being able to pay down that technical debt and handle that legacy. So I think, at least in my experience, the closer that I've gotten to talking in terms that are more concrete, of l... |
**Jon Sabados:** \[35:58\] Quantifying it can be interesting too, because yeah, there's some things you can be like "Yeah, this is a huge thing that's going to bite us hard." But when you're trying to quantify in exact terms for like "This may or may not be a problem, depending on what we do", which oftentimes maybe th... |
**Misha Avrekh:** Would it be fair to say that that is exactly the different between legacy code and technical debt, where technical debt is something that needs to be repaid and that will potentially grow with time, whereas legacy code, as Dominic said, it might just sit there and work for years until anyone notices? |
**Kris Brandow:** I feel like legacy code is like the technical debt that we haven't paid down for so long that we kind of declare bankruptcy on it. And the way that people think about legacy code - it's that code that's just like... Whenever you make that declaration of a greenfield, you're like, "Okay, we're done wit... |
**Dominic St-Pierre:** To me it boils down to once the company is near reaching a point where it will affect their bottom line, and when you're not able to find any engineers to work at your company - this is terrible. You've already reached a certain point that you should never have passed, in my opinion. But yeah, be... |
The other point from just a minute ago - I think we might see a new kind of engineering type evolving... Because if you think about it, how many software are we creating these days, compared to 25 years, 30 years ago? It's incredible. There will not be enough of us to maintain every single piece of software that's bein... |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.