We arrived in Paris on the Friday via the Eurostar from London and spent a couple of days doing the usual tourist things (art galleries, Churches etc.) before turning up to the Theatre De Paris for our second dotGo conference on the Monday morning.
The doors opened a few minutes late (time spent usefully chatting with others in the queue) but the whole registration process was quick and I had a stylish black Parisian gopher t-shirt at the end of it. The venue as a plush chic theatre of red decor was more uplifting than many tech venues and we feasted on chocolate pastries and coffee. I was lucky enough to recognise the go-vim author Fatih Arslan and spoke with him about some of his future plans since I was quite keen for delve debugger support (hoped for next year after vim 8 support had settled down). I also snarfed a couple of his rare go-vim stickers. I felt guilty taking two since they were in short supply but I had the legitimate excuse that Sue my partner (and fellow programmer) would want one.
We took a box at on the first floor which overlooked the stage and settled down. The Master of Ceremonies introduced himself on stage and presented Le Gopher —a mysterious cloaked figure — who spoke with a booming voice and was unveiled to reveal a massive (well over average human size) gopher stuffed giant.
Dave Cheney spoke first about "First-class functions" and credited as inspiration a Rob Pike Blog:
https://commandcenter.blogspot.co.uk/2014/01/self-referential-functions-and-design.html
Basically passing functions to functions is something more usual in dynamic languages such as javascript than more static, compiled languages like Go. But it was still possible in Go.
Dave gave a number of examples:
https://github.com/davecheney/presentations/tree/master/do-not-fear-first-class-functions
But the one which really stuck in my head was that of a calculator. The initial version had a dispatch type function containing a large ugly case statement which listed each arithmetic operation (eg. addition etc.) but by passing each operation as parameters to this function it could be refactored to look nicer and allow new operations to be more easily added. He also wisely encouraged restraint in the use of first class functions.
Damian Gryski then spoke about memory allocations, caches and performance:
https://github.com/dgryski/talks/tree/master/dotgo-2016
He gave some suggestions for trying to fit data in caches but concluded by saying in most cases just using simpler algorithms was better.
Then followed Dmitri Shuralyov speaking about GopherJS — a compiler which converted Go to Javascript — which I found quite exciting. As a server-side programmer I liked the idea of being able to avoid javascript and write Go to run as Javascript and being able to access the DOM etc. He quoted the examples of the frontend and backend being able to share the same business logic code and validation code as an obvious advantage. He also showed an ugly example of call-back hell Javascript to do HTTP streaming which refactored into much cleaner Go.
He also showed us an amazing "Hover Demo" running both on desktop and inside browser:
https://dmitri.shuralyov.com/projects/Hover-Demo/
(takes a little time to load!)
Péter Szilágyi spoke about Ethereum — a blockchain (like Bitcoin) based system where you could run programs. They had Denial of Service (DoS) issues last month based on caching bugs but he was able to fix this with immutability in Go. Go didn't have direct support for immutability but strings (and not slices) were immutable. Also he was able to make structs immutable indirectly.
https://ethereum.karalabe.com/talks/2016-dotgo.html
This was one of my favourite talks and I resolved to try the Ethereum system out.
The buffet food at lunch was absolutely superb, much better than the average tech conference food (London Go Conference I'm thinking of you!) although oddly there were no paper plates so one had to balance mini-pizza on tissue paper. Sadly Sue got quite angry about the toilet facilities for women and I learnt that in the entire building only one toilet cubicle downstairs was available for women (and shared with disabled!). But over all I thought the conference organisation was well above average (and I've been to many tech conferences). Many of us, including myself, had our photos taken with "Le Gopher" up on stage during this break.
We then had four lightning talks. Simone Carletti told us that using Go to develop a cross-platform API had led to better API design for the other languages:
Lucas Clemente talked about Google's fast QUIC (HTTP/2.0 over UDP) and his implementation::
https://github.com/lucas-clemente/quic-go
Janaa Burcu Dogan gave an excellent description of useful flags to pass to the Go build chain tools.
http://go-talks.appspot.com/github.com/rakyll/talks/gcinspect/talk.slide#1
Elias Naur spoke about Gomobile and told us it was now possible to directly call Android Java and IOS Objective C from within Go.
One of the conference highlights was probably Kelsey Hightower talking about "Self-deploying Go apps". A lot of speakers had shown kubernetes in use but he had taken this one step further to create his Kargo library which allowed this functionality to be embedded within a Go app itself.
https://github.com/kelseyhightower/kargo
Kelsey was a particularly polished and amusing speaker and did some scary live deployment demos to "rock the temple" of the Demo Gods (who didn't seem to mind too much) and show how the world of many computers could be treated as just one computer.
Some more lightning talks followed. Javier Provecho Fernández told us how to produce smaller Go binaries by stripping the binaries. Using the UNIX strip was wrong (which explained to me why OpenBSD's strip on Go binaries had actually produced *larger* binaries when I tried it!) but rather you should pass "-ldflags" options to "go build" in order to strip. Also he mentioned UPX to compress binaries with the warning it tended to trigger Anti Virus warnings.
I needed caffeination at this point and my note taking on my mobile was less zealous than before but I remember a lightning talk about:
https://github.com/sbinet/pygo
and Rhys Hiltner from Twich spoke on using
https://golang.org/pkg/runtime/trace/
to debug a production issue and also spoke of finding a Garbage Collection bug in Go itself (which would be explained in a future Twich video!).
Katrina Owen gave a nice example of code refactoring a zombie game (example cunningly rewritten to hide the guilty!) using the Flocking Rule and, sorry it has to be said, some quite ugly slide fonts!
Brad Rydzewski talked about how to use plugins (again something more associated with dynamic languages than static). He had a nice way of wrapping calls to external binaries using "net/rpc" for both os/exec and "docker run" and hosting external plugins on github as a convenient registry and showed us his Drone CI system as an example:
https://github.com/drone/drone
Matthew Holt, the author of caddyserver, talked about using Go with ACME (which is Let's Encrypt API to automatically provision SSL certs). The talk was a good one, practical, useful and certainly was effective in showing how much easier things had become than manual provision. He also gave reviews of Go libraries to use ACME.
https://drive.google.com/file/d/0B-OxZ0LoeTvGVDJXTThUT1Zwa2M/view
But, and I admit to straying into crypto politics here, I did think it was a little uncritical of the existing SSL infrastructure using trusted 3rd parties (like issues with some of the root CAs!). Self signed can work with a web of trust.
The last talk was Robert Griesemer, one of the original Go authors, who gave a superb talk on "Prototype your design!" using the example of how he was prototyping a new feature (multi-dimensional slices useful for matrices) with minimal modification of the existing Go compiler. He certainly explained things very clearly given the complexity of the subject.
https://github.com/griesemer/dotGo2016
Basically he hacked the part of the system where the Abstract Syntax Tree (AST) was written in order to cleverly shoehorn some new additional syntax in. This meant, with relatively minimal effort, he could actually try out the new syntax to see how well it worked with a view to adding something similar in properly in future. He explained step by step, illustrating with node and link type diagrams, as if you were watching the process of his work each change of the AST to add an addition operation.
Once this was done he had enough of an implementation to experiment with the syntax in order to review the design. He pointed out that "prototyping raises design questions we didn't even know we should be asking".
He had the feature, of that subset of the clever who are able to teach, of explaining things so well, by pitching at a high level, that for a few minutes I thought I understood it all very clearly. Then I thought a bit more and realised this was just the tip of the iceberg and how much more there was to it.
Anyway a great insight into how Go is developed and extended (and an example of how other software should be extended) and a great finish to the day.