text stringlengths 0 1.82k |
|---|
**Blake Mizerany:** Yes, correct. |
**Erik St. Martin:** I'd love to hear how you've seen both the usage of the language and the demographic of the community change over time, because you've been around since... It was first released and there were people poking at it, until now, which is almost everything being done, the distributed systems world is bei... |
**Blake Mizerany:** Yeah. Early on when I first saw Go, I took notice as I was working at Heroku, where we were working on a lot of distributed systems, and the appeal to me was mostly because everything we had done up until that point at Heroku and continued to do after that was a lot of Ruby, and it wasn't really hol... |
In fact, my colleague at the time - he's also an OG - Keith Rarick and I really started playing with Go together. Just before it was released, we both went to one of the founders of Heroku and said, "We really probably should start looking into building a language for distributed systems", which was really dumb for a s... |
**Brian Ketelsen:** Timing is everything. |
**Blake Mizerany:** Yeah, so we started using it pretty heavily after that. It took Heroku a while to adopt it, but Keith and I definitely used it for a lot of stuff. |
**Carlisia Thompson:** These stories always fascinate me. When I found out that tests in Go were running so much faster than same-size apps in Ruby - because I also came from Ruby and Ruby on Rails - that was what got my interest. I'm wondering with you, when you looked at Go, what "looked at Go" means? Did you do a dr... |
**Blake Mizerany:** Well, we didn't have to do any real formal benchmarking. Writing a quick little *Hello World* in Go, compiling and running it was... I mean, just a few little pokes at it with a browser; if it was a web server it was an obvious night and day difference from anything that we had written in Ruby. |
\[04:08\] It was pretty obvious at that point that that was the direction that we wanted to go, to start using Go. |
**Erik St. Martin:** I wanna say at that time, in 2009, to just hit an endpoint in Ruby on Rails was still three-digit response time; it was like 100-115 milliseconds, so the dropdown to single digits meant the world. |
**Blake Mizerany:** Right, yeah. I was working on a small project at Heroku at the time when I really started to use Go, and one thing I noticed was that we needed some sort of a service discovery and we needed some configuration management, so Keith and I picked up the Chubby paper that Google had put out, and we work... |
We were able to crank out a basic multi-Paxos implementation relatively quickly. I remember after we did that I immediately reached out to Rob Pike, sent him an email and said "Hey, we built a multi-Paxos implementation in Go. We'd love it if you'd come take a look, let us know what you think." So he came by -- I don't... |
That really spoke to the power of the language. You can work on some pretty hard problems in it and come up with some reasonably simple solutions. The language really allows for that. |
**Brian Ketelsen:** I have a confession... It was Doozer that got me hooked into Go. I played with it when it first came out in 2009, but I didn't really see the vision for Go yet. Then I saw Doozer and it blew me away. I realized that we could build all kinds of crazy stuff with it, and that's actually how Erik and I ... |
**Erik St. Martin:** Yeah, Doozer was a lot of fun to play with in the beginning. For anybody who's not familiar with it, it's a lot like Zookeeper and Etcd and Consul of this generation. There was a lot of functionality that it had that Zookeeper didn't, which was one of the reasons that we liked it. I wanna say it ha... |
**Blake Mizerany:** Right, we used the persistent data structure to be able to track changes, so you could pick up where you left off during a disconnect or after a disconnect, which was pretty nice. |
**Brian Ketelsen:** Yeah, so you could start watching a path at a certain point in time, which was really cool. |
**Erik St. Martin:** It was awesome for us, because we were doing some service discovery stuff in distributed systems, and when you looked at that, if you did Zookeeper, you would have to resync periodically; if you lost your session, you'd have to reconcile the differences. And you see that mirrored now in things like... |
**Blake Mizerany:** Right. There's always that chance that you'll miss the window. People offline for too long will have to resync, but it's more of a performance optimization. But it's still pretty cool. |
**Erik St. Martin:** So with that project, how much Ruby baggage do you feel like you brought along? Because that kind of turns out to be the thing whenever you adopt a new language. |
**Blake Mizerany:** \[08:02\] Right... One of the first things Keith and I wrote was an assert package for the testing suite. Not long after that we implemented a small Sinatra-style router called Pat, and within probably three to four weeks after creating both of those we ditched them and just never used them again, a... |
I remember looking for certain things to exist in Go that I was used to having in Ruby and it took me a while to break that habit. Once I did, everything got better. |
**Erik St. Martin:** I feel like for some of the open source projects you release into the world, there's some of them I feel like I should go back and change the readme to be like, "I no longer use this, neither should you." |
**Blake Mizerany:** Yeah, I did. If you look at the Pat readme, I think it literally says "I don't use this anymore." \[laughter\] |
**Brian Ketelsen:** One of the first questions that almost everybody asks when they come to Go is "Which web framework should I use? Which framework should I use for this? Which framework should I use for that?" and it's really a strong question that we get, especially in the Gopher Slack channel, every day. I know you... |
**Blake Mizerany:** Yeah... I don't really have a go-to answer for that. I think everyone has their own goals when they come to the language. It really depends on what they're trying to do. If someone wants to write a simple web app, they can certainly get away with just using the standard library. The standard library... |
I think some people are also looking for, you know, "I've got a checklist of things I need out of the box so that I can get started today", and I don't have an answer for them on that, but I hear things about Gorilla or Martini or things like that. To answer your question... I don't know, I don't have an answer. \[laug... |
**Carlisia Thompson:** I don't know how it is with the Python community or other communities, but with the Ruby community the message was so strong, "Don't reinvent the wheel. Don't write things yourself if there is a library out there. Use it." Also the message of "Extract your code into a gem and use that, and also s... |
**Blake Mizerany:** Right. Usually what I advocate in those scenarios is I go look for something that does what you're looking for and then copy and paste it into your code. Obviously, you can give attribution as well, but usually what I see when people ask that is that they're looking for a quick fix, for something th... |
\[12:07\] Not to point fingers to anyone, but I've seen libraries where people have written an entire library to do exponential backoff. To me, that's a really simple four-loop and a random sleep; it's nothing too complicated. What I see in those libraries also is that what they wind up doing is you wind up with someth... |
**Carlisia Thompson:** That is a very good point. I started writing a greenfield app that I know is going to be in production sometime soon, and of course my being new to Go, my first reaction was "Well, let me look for some libraries that do this stuff that I haven't done, since I worked with Java." I haven't done cod... |
And like you said, I think your point is so good, because it ends up not being that hard. Now, after I looked at it, I was like, "Okay, this is not that complicated." |
**Blake Mizerany:** Right. |
**Erik St. Martin:** I think you hit on it a little bit at a talk you did - I believe it was a DotGo a couple years ago... You talked about pre-determinism. When you attack a problem and you think you already know what you need, and for most of us, especially when you come from Python, Django or Ruby with Rails, you fe... |
**Blake Mizerany:** On the other side of that, I wouldn't recommend implementing your own database driver, or cryptography library, or bucketed rate limiting. Those types of things aren't something you just kind of throw together pretty quickly and then be done with it; they require a little bit of thought. The nice th... |
**Erik St. Martin:** That's a fair point. |
**Blake Mizerany:** That's generally when I start bringing things in, and bringing dependencies in is when it's something like that. |
**Erik St. Martin:** If it's a problem that is only ever explained in formal papers, you probably want a library for it. |
**Blake Mizerany:** Right. I implemented the first Postgres driver for Go for the database SQL package, and I'm here to tell you that it's not something you just wanna do. I had to do it at the time, because nothing existed, but I would have preferred someone else did it. |
**Erik St. Martin:** \[16:08\] That's a good question, too... Because you came in so early and there weren't a lot of libraries existing already, so you kind of had to pave the way anyway. Do you feel like that kind of influences your desire to just kind of write it from scratch most of the time? Or is it really just y... |
**Blake Mizerany:** It depends on what we're trying to do... When I say "we", I mean us at Backplane. We're trying to get something done by the end of the week; there's no problem with bringing a dependency in, but we'll definitely reevaluate it pretty quickly. But there is this strong urge a lot of times where I just ... |
**Erik St. Martin:** Yeah, I definitely think it does. There's a "not invented here" syndrome that people can get wrapped up into, but like you said, you just kind of have to be pragmatic about your decision of whether or not to pull in a dependency or rewrite it yourself. |
**Blake Mizerany:** And be self-aware as well, just to make sure you're not doing it just to do it. |
**Erik St. Martin:** Unless you want to, for fun. |
**Blake Mizerany:** Right, exactly. There is some context needed there. If you want to do it for fun, then by all means, you're trying to get something out the door, put something into production. |
**Brian Ketelsen:** I know we've only got a few minutes left with you because you've got a hard stop in a few minutes here, but my big question for you is around building a startup in Go. What advice do you have for other entrepreneurs who are considering using Go for startups? |
**Blake Mizerany:** I think they should, but again, also depending on what it is they're trying to do. For us, it's been a real boom, we're extremely productive. We're working on a lot of distributed systems-y things here, and Go suits us extremely well and helps solve a lot of the problems with relative easy, while al... |
**Brian Ketelsen:** No, I was just curious, now that you've had almost a year in Go as a CEO of a startup, is there anything that you would change? Are you having a difficult time finding developers? Is the talent pool big, is the language mature enough? Those are all questions that investors and entrepreneurs want to ... |
**Blake Mizerany:** Right. I think what's great about Go is that it's easy to learn, so if you're having trouble finding people that aren't as senior as you might hope, the great thing about it is that people can learn extremely fast. I did spend quite a bit of time - probably I spent too much time early on trying to f... |
**Brian Ketelsen:** \[20:18\] Very good. Alright, we know that you've gotta go quickly, so we will say our thanks and goodbyes. We appreciate you coming on the show today. |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.