←back to thread

Alan Kay has agreed to do an AMA today

1401 points alankay | 1 comments | 20 Jun 16 18:04 UTC | HN request time: 0.223s | source

This request originated via recent discussions on HN, and the forming of HARC! at YC Research. I'll be around for most of the day today (though the early evening).
Show context
nostrademons ◴[20 Jun 16 18:15 UTC] No.11939944[source]▶
What turning points in the history of computing (products that won in the marketplace, inventions that were ignored, technical decisions where the individual/company/committee could've explored a different alternative, etc.) do you wish had gone another way?
replies(1): >>11945435 #
alankay ◴[21 Jun 16 13:32 UTC] No.11945435[source]▶
Just to pick three (and maybe not even at the top of my list if I were to write it and sort it), are

(a) Intel and Motorola, etc. getting really interested in the Parc HW architectures that allowed Very High Level Languages to be efficiently implemented. Not having this in the 80s brought "not very good ideas from the 50s and 60s" back into programming, and was one of the big factors in:

(b) the huge propensity of "we know how to program" etc., that was the other big factor preventing the best software practices from the 70s from being the start of much better programming, operating systems, etc. in the 1980s, rather the reversion to weak methods (from which we really haven't recovered).

(c) The use of "best ideas about destiny of computing" e.g. in the ARPA community, rather than weak gestures e.g. the really poorly conceived WWW vs the really important and needed ideas of Engelbart.

replies(1): >>11957106 #
jonathanlocke ◴[22 Jun 16 21:19 UTC] No.11957106[source]▶
I get (a) and (b) completely. On (c), I felt this way about NCSA Mosaic in 1993 when I first saw it and I'm relieved to hear you say this because although I definitely misunderstood a major technology shift for a few years, maybe I wasn't wrong in my initial reaction that it was stupid.
replies(1): >>11957620 #
mmiller ◴[22 Jun 16 22:43 UTC] No.11957620[source]▶
I didn't begin to get it until the industry started trying to use browsers for applications in the late '90s/early 2000's. I took one look at the "stateful" architecture they were trying to use, and I said to myself, "This is a hack." I learned shortly thereafter about criticism of it saying the same thing, "This is an attempt to impose statefulness on an inherently stateless architecture." I kept wondering why the industry wasn't using X11, which already had the ability to carry out full GUI interactions remotely. Why reject a real-time interactive architecture that's designed for network use for one that insisted on page refreshes to update the display? The whole thing felt like a step backward. The point where it clobbered me over the head was when I tried to use a web application framework to make a complex web form application work. I got it to work, and the customer was very pleased, but I was ashamed of the code I wrote, because I felt like I had to write it like I was a contortionist. I was fortunate in that I'd had prior experience with other platforms where the architecture was more sane, so that I didn't think this was a "good design." After that experience, I left the industry. I've been trying to segue into a different, more sane way of working with computers since. I don't think any of my past experience really qualifies, with the exception of some small aspects and experiences. The key is not to get discouraged once you've witnessed works that put your own to shame, but to realize that the difference in quality matters, that it was done by people rather like yourself who had the opportunity to put focus and attention on it, and that one should aspire to meet or exceed it, because anything else is a waste of time.
replies(2): >>11959400 #>>12010226 #
1. jonathanlocke ◴[30 Jun 16 17:34 UTC] No.12010226[source]▶
I don't think networked X11 is quite the web we'd want (it's really outdated), but it does seem better than browsers, which as you point out are so bad you want to stab your eyes out. Unfortunately, now that the web has scaled up to this enormous size, people can't un-see it and it does seem like it's seriously polluted our thinking about how the Internet should interact with end users.

Maybe the trick is something close to this: we need an Internet where it's very easy to do not only WYSIWYG document composition and publishing (which is what the web originally was, minus the WYSIWYG), but really deliver any kind of user experience we want (like VR, for example). It should be based on a network OS (an abstract, extensible microkernel on steroids) where user experiences of the network are actually programs with their own microkernel systems (sort of like an updated take on postscript). The network OS can security check the interpreters and quota and deal out resources and the microkernels that deliver user experiences like documents can be updated as what we want to do changes over time. I think we'd have something more in this direction (although I'm sure I missed any number of obvious problems) if we were to actually pass Alan Kay's OS-101 class as an industry.

We actually sort of very briefly started heading in this direction with Marimba's "Castanet" back at the beginning of Java and I was WILDLY excited to see us trying something less dumb than the browser. Unfortunately, it would seem that economic pressures pushed Marimba into becoming a software deployment provider, which is really not what I think they were originally trying to do. Castanet should have become the OS of the web. I think Java still has the potential to create something much better than the web because a ubiquitous and very mature virtual machine is a very powerful thing, but I don't see anyone trying go there. There's this mentality of "nobody would install something better." And yet we installed Netscape and even IE...

BTW, I do think the security problems of running untrusted code are potentially solvable (at least so much as any network security problems are) using a proper messaging microkernel architecture with the trusted resource-accessing code running in one process and the untrusted code running in another. The problem with the Java sandbox (so far as I understand all that) is that it's in-process. The scary code runs with the trusted code. In theory, Java is controlled enough to protect us from the scary code, but in practice, people are really smart and one tiny screw-up in the JVM or the JDK and bad code gets permissions it shouldn't have. A lot of these errors could be controlled or eliminated by separating the trusted code from the untrusted code as in Windows NT (even if only by making the protocol for resource permissions really clear).