5,495 karma · joined September 17, 2008
http://www.storytotell.org
Feel free to email me at fusion at storytotell dot org.
[ my public key: https://keybase.io/fusiongyro; my proof: https://keybase.io/fusiongyro/sigs/aYrkbCZouBBhqp_qOKPGdBfIn---A8P8wPVCKLSlpnA ]
I don't think awareness of Lisp was evenly distributed among people developing on PCs in the time period when the network advantage was obtained. Even if it was, if there were already a decent C compiler, I don't think most practical engineers would immediately embark on a compiler development project. During this era, those who did know about Lisp certainly would have regarded it as being too resource-intensive for work in a resource-constrained environment like the PC.
I think it's specious to say it's easy to develop a simple Lisp compiler, so why didn't anything get ported. The code you may have wanted to port would not have been written in some minimalist dialect of Lisp, it would have been written in ZetaLisp. So there would be assumptions about resources and the underlying system that you would have to fake. And again, we're talking about porting from a resource-rich environment to a resource-scarce environment.
It was probably a reach for me to say there wasn't anything to port. But what would someone have wanted to port? A graphical application that ran on a Lisp machine would have made assumptions about the display that aren't true about a PC. An academic AI program would have been too short on resources to be much use.
What wound up selling the PCs in the early days was business applications. Were there an assortment of useful business applications on Lisp machines that there was commercial demand for? WordStar and Visicalc invented that territory and were first released in the late 70s—does anyone remember programs for the Lisp machine that businesses were clamoring for? And really, software churn has always meant there are more new programs in use than old ones.
What I don't have an explanation for is why the C compilers produced better code than the Pascal compilers. Partly, I know p-code made it faster to build a Pascal compiler but I don't know why it inhibited development of more performant ones. Turbo Pascal's speed of compilation was a result of not having an AST, which probably inhibited performance optimizations, but that may have been too late in the story to explain anything. And I don't have an explanation for why nobody developed a Lisp-like language that had explicit allocation. My best guess would be simply that their attention was elsewhere, probably working on AI, which wouldn't bottom out until after the winners were already entrenched.
By the way, in case it isn't clear, I am really enjoying this conversation. I'm very interested in this window of time in computing and if I have a wrong idea about something, I do want to know. The only way to preserve this history is to talk about it and keep the memory alive.
I'm not denigrating Lisp. I'm saying the network effects favored C because the PC happened to take off.
Lisp had a head start, but wasn't on the track that mattered.
It's not a very meaningful answer to the question, but it's a bit like asking why Todd isn't rich when we all know Todd is a smart guy. It's because getting rich takes more than raw intelligence.
Suppose you have a wildly evolving field where nobody can really predict what technologies are going to be successful in the market beyond the next five years or so. Suppose you have no internet. Suppose there are thousands of programmers with the ability to invent a language. Wouldn't you expect that some of them will, by happenstance, wind up on successful products and become successful while others stay relatively unknown? That's the birth of the mainstream programming languages and their antecedents in the 70s and early 80s.
https://hanshuebner.github.io/lmman/files.xml#fascism
I think there are a few other jokes in the index but I don't recall what they are.
Many of the NoSQL databases have provisions for these situations so they can handle both writes but bubble the decision up to the application layer. CouchDB, for instance, will tell you you've got multiple versions and has a way for you to tell it how to resolve the conflict. CRDTs let you bake the logic in for handling a conflict. MySQL does not do any of that. If your database consists of "likes" and other low-value data like that, you may not care if you're losing data at some low rate. Reddit is in that situation, and the decision seems to work for them. But I don't think it generalizes to all situations.
People my age and younger just didn't get much access to commercial Unix environments when we were growing up. Once we got into industry, we found it hard to see the justification for high Unixes when we had experienced good results with low Unixes. So the low Unixes have mostly won the war. This story repeats time and time again in our industry: the freest foo conquers the non-free foo, even if the non-free foo is technically superior.
There are certainly features MySQL has that Postgres doesn't, but data integrity is problem 1, and most of us got burned by MySQL on that one at some point before we got here. Or, we deployed MySQL and then found out that the features we were depending on were InnoDB-only, but switching to InnoDB cratered our performance. Or we tripped on one too many of MySQL's numerous special cases or silent limitations.
You can use MySQL productively. You can even benefit from its special features. But the thing taken together is a bit of a mine field.
Postgres has some false advertising issues too ("object-relational"?) but the day-to-day living with it is a lot nicer, because the roof doesn't leak. I can live without the granite counters.
I think Postgres will be a better product when it incorporates similar functionality, but the implementations in MySQL and Oracle lie somewhere on the border between false advertising and reckless endangerment. I appreciate that Postgres not having it as two-lines of configuration sends a message that you are making a commitment. Nearly everything else in Postgres is set-it-and-forget-it. You really want to have a DBA if you're going to try to make replication work. When you read the docs on having a hot standby and how much of a commitment that is and sense yourself getting tired, you're starting to appreciate what you might be getting yourself into with single-master replication, let alone multi-master.
It was difficult to get much experience with Solaris or the other commercial Unixes because you usually had to have particular hardware. Even when Solaris became free and ran on x86, it was picky about hardware. By then it seemed to me that even running BSD was kind of an uphill battle, if you wanted a VPS or something.
I'm not a completely neutral party. Sun totally botched a job interview process with me when I graduated, and I came away feeling that there must be a lot of internal confusion. Their public side certainly gave that impression too. I knew they had cool technology in Solaris, like dtrace and zones, but apart from the names you almost never saw someone who attributed their success to these technologies. And then of course Oracle and the slow death.
Another way to look at is that digital audio is not like digital imaging. There aren't pixels. Increasing the data rate does not continue to make the waveform more and more detailed to human auditory perception in the way that raising the pixel density does for human visual perception.
To describe it intuitively, forget your intuition that audio is like visual and start from "there is no metaphor between these two things."
I have been joking about how we should build a "total sh*t array" out of old Dish/DirecTV antennas, so that we could explore the systems design without worrying too much about whether anything could be done with the data collected. This hasn't interested my coworkers that much :) There is an amateur radio astronomy society, and there are plans for how to build various levels of radio telescope, starting from ~$50 and an old Dish receiver and going up. And our open-skies policy means that you do not have to be a professional astronomer to use our instruments, although I only know of one or two amateurs that have proposed for time. (They did get it, though).
As a rough back-of-the-hand deal, we allow anybody who gets time to access about 25 MB/s of data. The correlator we have currently (WIDAR) can certainly output much more than that, up to gigabytes per second, but the rest of the infrastructure certainly can't keep up at that rate sustained. It's not unusual for an observation to top a few TB in size. ALMA data files are probably even larger on average.
We are already in the early design stage for a next-generation VLA, which will increase the number of antennas to about 300. At that point, we probably won't be able to keep correlated but unprocessed raw data, just because of the sheer size of it.
ZeroMQ is getting used more for those kinds of purposes; the Greenbank Telescope uses it for one of their instrument backends and we are now using it for VLITE and REALfast. The new archive system I'm helping build uses AMQP.
https://racket-lang.org/new-name.html
A lot of modern languages seem to eschew what they are derived from; Pharo isn't Smalltalk either, SWI-Prolog now defaults to a standards-incompatible mode, etc.
I also think it would be interesting, albeit in a totally different way, to know what "algorithms" they're up to in there, even though it's bullshit.
Definitely agree about their quackery.
There's a great book called Practical Astronomy with your Calculator; I wonder what the astrological analog would be.