And if your language of choice doesn't have a CBOR lib, CBOR is fairly easy to implement and writing a encoder/decoder is very fun! I recently completed my implementation for the Gerbil Scheme language last week [0].

[0]: https://github.com/chiefnoah/gerbil-cbor

But now the interoperable serialization formats have a lot more energy spent on tools and such, and the formats are better-specified, to the point that you probably want to use them even where interoperability with other stuff doesn't force it.

The thing is that at Google protobufs are used everywhere. Like, absolutely everywhere. Think of all the places they could be, and it's way more. All the tooling understands them, code search and go-to-reference works on them everywhere, they are truly transformational on how many different services with many different implementations interact.

Are they perfect? Far from it. But a Go-only implementation misses almost all the value of protobufs. If I was inventing them for Python I could do better (pickle? maybe not) but the whole point is that they aren't language/ecosystem/use-case specific. If this was Rob Pike's frustration then I can't help but feel he missed the point, or this post is a little disingenuous as to the benefits.

I've not seen Gobs used in Google, but I'd imagine an engineer would need to make a very strong case against using protos between services, regardless of if both services are in Go.

Did he miss the point, or was the point to light a fire under the protobuf team to improve their product? Which they eventually did (e.g. required fields were removed after this post was made).

Back when he did his best work, it was possible for one person to "just write the new thing", without making it fit with anything else. There was nothing else to fit it with. You could invent everything from scratch, and not only was it not a waste of time, if you were good enough it had a chance of being the best fit for purpose.

You could take shortcuts. You could have every part of your system be "odd", because nothing was "even".

That's not true anymore. And the way I see it Pike has not moved on.

Science in general had this switch at one point, too. There was a point where one could know all of science. But it's long gone.

gob was his opinionated way of doing a Go-specific encoding, while also supporting any number of other encodings in the language. Go has incredibly good support for almost every popular encoding there is.

gob has also been used successfully by a number of projects. In many cases, it's a perfectly good way to encode a piece of data that is completely local to a Go program.

I didn't mean it as such.

> in the midst of doing his best work creating Go itself.

Not to go too far off topic, but Go is another example. It famously ignores decades of language theory, and they wrote their own assembler, linker, etc.

Now, much of that has been undone and rewritten, as Go became more adopted, requiring playing well with the rest of the ecosystem.

(but much of it we're unfortunately stuck with, because it's part of the language)

30 years ago there was no ecosystem to play well with, and compared to now we were just banging rocks together. Back then you could be a CS polymath as one person. Well, I couldn't, but Pike could.

It was the old days of John Carmack starting every game engine with an empty directory.

I'm saying that today nobody can. Even John Carmack could not on his own write a AAA game. (I know ID had other coders, but my point stands)

The idea that they did so out of ignorance is ridiculous, given the background of the Go design team. They made considered decisions about what to include in Go.

A fixed-gear bicycle is not "ignoring" decades of bicycle design theory.

Creating a programming language is an engineering task not a theoretical task. Which means there are major trade-offs to be made. And they chose their trade-offs. The wild success of Go should at least make you consider whether or not they're better at making these trade-offs than you, and most language designers, are.

Maybe they were wrong about some choices, which is why Go is still evolving, but they were self-evidently mostly correct.

You seem to have gotten a bit emotional about this, so I don't think this'll go anywhere.

Claiming the poster is being emotional is... poor form, to put it mildly

Putting scare quotes around what I said, calling what I said "ridiculous", and seems to have taken critique of the Go language personally by saying that the Go language designers better than most language designers.

So this is why I found the reply neither calm nor courteous. As for reasoned, I don't see any reasoning it it, just conclusions.

I don't even see any sign of a refutation of my point, which isn't so much about if Go is good or not, but about the needless reinventing and ignoring other work, and as a result running into problems that would have been predictable, had they been taken into account.

To me it summarizes as "they're smart, you're dumb".

If this is the type of replies this person would write, then that's not productive.

I've seen Pike have conversations about language design with SPJ, Hejlsberg, Lattner, Stroustrup, Odersky and other highly respective PL designers and they would never make such a shallow and trite comment.

True, but over 30 years ago. In fact, said compiler was the very first program written for Plan 9. Those decades you speak of came afterwards.

Disclaimer: an employee at the said moving company.

By the way, just curious what is your opinion on MessagePack?