Rob Pike has apparently never had an internet outage, or traveled outside modern civilization?
Independence is a virtue. Even in the old phones-as-infrastructure case, there were many times when there was simply no working phone around, or available. Broken down on the road, being in an industrial area of town on the weekend, etc.
If everything your life revolves around requires working and accessible infrastructure to access it, you're going to have times where you're left staring at a connection error screen, or no screen at all. And because the universe hates you, that will happen at the worst possible times. ;-)
He's at Google. Many people there seem to think of network connectivity the same way we think of clean water, electricity, and natural gas - things that are just always there, in whatever quantity you want. Ditto for storage and computational power, because for them all those things are like utilities. Never mind that people in many parts of the world can't even take those "old fashioned" utilities for granted. People who live in bubbles tend to lose touch with the outside world.
Thinks like the Andrew File System and Coda were actually developed to try and work around this, to provide resilience and allow one to have "soft" state in the presence of outages. The interest was not there, certainly not when most people were working on standalone Windows boxes or Macs, and didn't have any idea of the sort of power that was available in research labs via X and Ethernet.
It's a shame, because not only did research groups think about these issues, 20 solid years ago, but also several strong contenders for solutions were implemented. Society did not see the value in supporting further work on the projects and so the researchers moved on.
How is it "pragmatic" to assume infinite connectivity instead of working to solve the problems when that assumption fails? That word does not mean what you want it to mean. I've worked on those problems for a couple of decades too. I also bemoan the relative lack of attention given to them. That's exactly why I find his attitude so disappointing.
It's pragmatic to delegate that aspect to other people who are/were working on it. It is not pragmatic to try and solve all the different problems involved all at the same time.
Pike was brought in to Google to do something about log management (hence Sawmill -- it was a disaster prior to his arrival). The solution was pragmatic: awk on GFS.
It would be nice if more attention to intermittent connectivity solved the problem. The trouble is that I see no interest from research funders and not much motive for commercial funders. As best as I can tell, it is in fact quite pragmatic to leave that detail (important as it may be) aside until there's enough interest in stateless "dumb" workstations to reintroduce it.
It's pragmatic to make flawed assumptions so that you can get something done in the common-within-Google case, and come back later for the common-outside-of-Google case, at least if you happen to be within Google.
That would be fine if he had been talking about what he was doing within Google, but he was specifically talking about his dream system. Here's an analogy.
"I have a dream about a programming language and runtime that will automatically determine how serial code can be distributed across a bunch of machines, and do so without totally screwing up performance or failure semantics. But that's somebody else's problem. I'm not even going to mention any actual ideas about how to do it."
Would you say "pragmatic" is the right word to describe that? I wouldn't. We left the realm of pragmatism as soon as we started talking about dreams. It's not necessarily wrong for me to mention such a dream. Such a compiler would make my life a lot easier. However, it misses an opportunity to inspire would-be collaborators in areas where I have concrete ideas and actually am working to make a difference. It contributes less to actual progress than another answer might have, and that's disappointing.
The thing about people like Pike, from my perspective: When that scope changes, they tend to be effective at addressing this and coming up with practical, efficient solutions.
I don't expect better from him than from myself: Tend to the problems at hand.
In fact, I too often get too lost in the "big picture". Maybe one reason I don't have a comparable level of achievement.
Not that Pike doesn't see the big picture. Maybe he just chunks it better. :-)
Independence is a virtue. Even in the old phones-as-infrastructure case, there were many times when there was simply no working phone around, or available. Broken down on the road, being in an industrial area of town on the weekend, etc.
If everything your life revolves around requires working and accessible infrastructure to access it, you're going to have times where you're left staring at a connection error screen, or no screen at all. And because the universe hates you, that will happen at the worst possible times. ;-)