Hacker News Reader

Comment by slowin | Hacker News Reader

A lot of people don't know that Smalltalk heavily influenced NeXTSTEP and Objective-C. You'll see a lot of it in Interface Builder->Project Builder->XCode with the GUI serialization (and of course Obj-C message passing and general dynamic nature).

This is a good history:

https://computerhistory.org/blog/the-deep-history-of-your-ap...

View on HN

Which then influenced Mesa, Cedar, Interlisp-D inside Xerox.

Mesa and Cedar influenced Niklaus Wirth to come up with Modula-2/Lillith and Oberon.

Which had an influence on Rob Pike, on how he developed ACME for Plan 9.

Meanwhile StrongTalk, and Objective-C ended up influencing how Java came to be, and its approach to JIT compilation.

On another track Smalltalk ends up influencing Self design, which ends up having an effect on how V8 JIT was designed.

Then there is Ruby as well.

Pity that with all of this, many still never tried what actually means to use Smalltalk.

Note that Interface Builder started as a Lisp prototype, before being acquired by NeXT.

https://paulhammant.com/2013/03/28/interface-builders-altern...

The LeLisp version of Interface Builder that JMH wrote at INRIA predates the timeline in your link.

That link is the official story I am aware of, where can I read about that one?

I have a paper copy of a paper (in French) about SOS Interface. Had scans of this at one point.

There is a bit more of a timeline in a talk [1] he gave at INRIA a couple of years before he died.

[1] https://www.inria.fr/en/jean-marie-hullot-perforated-cards-i...

Thanks.

> Which then influenced Mesa, Cedar

In what respect? What sources for this claim do you have?

The Xerox papers on digital archive, I can try to hunt them down, see for example the one on XDE Tajo experience for Mesa.

MPS/MPL began in 1971; the Mesa name was adopted in June 1974; an Alto cross-compiler worked in spring 1975. Cedar is both a language (an extension of Mesa, e.g. by a GC, but the OO features proposed by Lampson were never implemented to my knowledge) and a development environment. I assume your claim is about the latter. The programming language doesn't have anything in common with Smalltalk, and the fundamental work even predated the latter.

Yes my claim is about the IDE experience shared across all Xerox PARC environments, and how it affected the design of each language individually.

I will hunt down the papers.

We should not forget that the original Alto operating system(s) had a GUI already and was the basis of well known applications like Bravo or Gypsy; the original OS and the said applications were implemente in BCPL; and all were operational before Smalltalk-76. By Wirth's first PARC visit in 1976–77, Mesa had already become a self-hosting systems-development tool (small operating system and symbolic debugger); I'm not aware of any substantial Smalltalk influence at that time.

> I'm not aware of any substantial Smalltalk influence at that time.

You should be, because I spent hours telling you about it, in detail, with citations and extensive quotes from the designers of the Alto and Bravo, less than two weeks ago. At the time you conceded the point and said it was not what you intended to dispute: https://news.ycombinator.com/item?id=49857855

> I spent hours telling you about it

I hope not.

And please note that this is again a different topic.

Last time I was talking about Wirth's OO vs. Simula OO, and the similarity of Wirth's implementation with Kay's vision. You then first extended this discussion to whether Kay directly influenced Wirth, which I responded. Then you extended the topic even more to the question whether and to what degree Cedar was influenced by Smalltalk. That was far off my interest and irrelevant for the discussion given that Wirth just took the tiling window manager, which none of the Smalltalk versions had.

Now I'm talking about the development of the Alto computer which included an OS with a GUI and two groundbreaking applications, Bravo and Gypsy (implemented on top of Bravo), all in BCPL, implementing their own interaction and presentation concepts.

If you want to do historical research, you have to go to the primary sources (e.g. https://www.mirrorservice.org/sites/www.bitsavers.org/pdf/xe..., https://xeroxparcarchive.computerhistory.org/Xerox_PARC_sour..., or e.g. https://archive.org/details/Xerox_PARC_Gypsy_Demonstration_T...), not to stories written down from memory twenty years later.

The reason the Alto included a GUI at all was specifically Alan Kay's vision, as explained in the text I quoted by the people who built it. That text includes primary sources written at the time by, among other people, Cedar's main author and Bravo's main author — not only, as you imply, stories written down from memory twenty years later. It also includes primary sources from many years later, but those add to the evidence; they do not subtract from it.

I do not believe that Wirth's "OO" is a different topic. To the best of my understanding, the reason he added procedure-typed record fields in Modula-2 and retained them in Oberon was specifically in order to implement that tiling window manager, which in later versions of Mesa (and Cedar) replaced the overlapping-window GUI in earlier versions of Mesa, which came from Alan Kay's 01969 doctoral dissertation, by way of Smalltalk.

Were you two having a discussion about extensible programming languages, or an extensible discussion about programming languages? Or was it an extensible discussion about extensible programming languages after all? Maybe it all boils down to whether discussions about programming languages should use static or late binding?

Rochus has posted a response saying:

> Kay's 1969 thesis has no overlapping windows; its "windows" are Sutherland–Sproull clipping viewports into a vector-display virtual screen ("the terminology follows [A Clipping Divider, 1968]", p. 40); nothing occludes or moves. Overlapping bitmap windows first appear in Smalltalk-72 on the Alto (1973–74), implemented by Ingalls. Tiling did neither come from Mesa nor from Kay; multiple tiled windows were on screen in Engelbart's NLS in 1968 (Engelbart & English, FJCC 1968, and the demo), carried into PARC by his former team (English, Irby), and are the stated model for the Mesa debugger. Procedure types in Modula-2 (1977–80, for Lilith) were taken from Mesa after Wirth's 1976–77 PARC stay, years before Cedar's tiled Viewers (1982), which Wirth saw in 1984–85.

Unfortunately, and horrifyingly, this response is [dead].

I agree with some things in it, but not others, and I will have to write a longer response later.

> Unfortunately, and horrifyingly, this response is [dead].

That's strange. I didn't do anything and it still appears on the site in my browser. Did they add "shadow bans" to HN? That was one reason for me to turn my back on Facebook (and the fact that interesting discussions suddenly disappeared later on). I took me more than half an hour to re-check all sources; I (we) probably shouldn't spend that much time on comments anymore.

There are shadowbans on HN, but you are not shadowbanned. Some assholes just flagged your comment, that's all. I appreciate the comment, and will respond to it in due course — I don't want to depend on my own memories here, but also to re-check the sources myself.

Here is one such paper,

https://www.lri.fr/~mbl/ENS/FundHCI/2013/papers/XeroxStar-Co...

A few quotes from it,

> SunView is an example of a window system that is based upon the Tools metaphor rather than the Desktop metaphor. Its users see a collection of application program windows, each of which is used to edit certain files. Smalltalk-80, Cedar, and various Lisp environments also use the Tools metaphor rather than the Desktop metaphor

> The Reactive Engine -- While Engelbart et al were developing ideas, some of which eventually found their way into Star, Alan Kay, then a graduate student, was doing likewise. His dissertation, The Reactive Engine, contained seeds of many ideas that he and others later brought to fruition in the Smalltalk language and programming environment, which, in turn, influenced Star.

> Most importantly for Star, they demonstrated the power of graphical, bitmapped displays, mouse-driven input, windows, and simultaneous applications. This is the most visible link between Smalltalk and Star, and is perhaps why many people wrongly believe that Star was written in Smalltalk.

> Even though the machine was designed to run Star, it also ran other software. In addition to selling it as the 8010 "Star" workstation, Xerox sold it as a server machine and as an Interlisp and a Smalltalk workstation.

From this other one, some quotes as well,

"The Mesa Programming Environment"

https://dl.acm.org/doi/pdf/10.1145/17919.806843

> Client programs can attach additional information to windows by means of named contexts, similar to property lists of LISP. Tools can be written to have multiple instances, obtaining their data from the context associated with the particular window that Tajo is notifying (passed as a parameter to the notify procedure).

> The debugger had multiple windows patterned after those of Smalltalk [Kay76]. One, a typescript window, was used for interaction with the debugger executive for setting breakpoints, examining data, etc., and another was used to show the source context at breakpoints.

There are a few others with similar one off sentences like these, telling how designers from one system got influenced by the other, while at Xerox PARC.

Thanks. Your quotes are not about the Alto system software. The Star paper is about Star (1981) and itself warns that people "wrongly believe that Star was written in Smalltalk"; your first Sweet quote credits Tajo's window contexts to Lisp property lists, not Smalltalk.

The Alto OS, Bravo (1974) and Gypsy (1975) were written in BCPL by CSL/POLOS people (Lampson, Sproull, Simonyi, Tesler, Mott); the June 1975 Alto OS manual and Alto Packages contain nothing from Smalltalk, and they had no overlapping windows or pop-up menus.

Mesa the language (MPS 1971, named 1974, self-hosting 1976) owes nothing to Smalltalk either. Sweet's "patterned after those of Smalltalk [Kay76]" refers to the Dec 1976 Mesa 1.0 debugger's optional, "experimental" window manager. It never touched compiler, binder or OS and was dropped/replaced in Mesa 4.0.

Smalltalk influence is documented for Star and for Cedar's environment goals, both 1980s. For the Alto software of 1973–75 there is none; for Mesa 1976–77 it is BitBlt plus an experimental debugger add-on. BitBlt and overlapping windows were specifically Ingalls' work, which he internally demonstrated in 1975.

Paul McJones has a nice Mesa History Collection with a lot of primary sources from the time: https://softwarepreservation.computerhistory.org/mesa

You and I already went through this discussion at great length two weeks ago, and I came up with lots of sources: https://news.ycombinator.com/item?id=49857855

Mesa and Cedar are both programming languages which have nothing in common with Smalltalk. Cedar is also a later development environment. The original post was about Smalltalk vs. Objective-C, i.e. a language comparison, and then mentioning some graphic tools as well. Thus the comment I responded to was ambiguous in different respects. That's why I asked. It has nothing to do with the discussion you referenced.

Mesa and Cedar interested me as I got into programming languages, but I was more into Trellis/Owl (and had copies of the documentation provided to me by DEC SRC after I wrote to them).

Trellis/Owl, now that's a trip down memory lane. Kilian's paper on the compiler was my entry point into doing databases for the last ~30 years, for odd reasons.

I really don’t remember much influence of smalltalk on interlisp at the time (I worked in the KSL, though I was working on 3Lisp so Interlisp was in theory just our implementation language…but of course boundaries were not what PARC was about.

The development experience while using the environment, the debugger, edit-compile-continue, REPL with suggestions for typos, dynamic code loading.

I should made myself clear it wasn't the programming languages grammar or semantics by themselves.

All of that was common in MACLISP (MIT) and InterLISP (BBN) before PARC was founded. It was part of the lineage Alan talked about when founding PARC. DWIM was Teitelman’s PhD thesis (BTW his advisor was Minsky, which seems odd today but was normal back then).

Rochus and I were arguing about this a couple of weeks ago, and I read through a bunch of historical papers, including Teitelbaum talking about how Smalltalk influenced Interlisp-D and Lampson more or less attributing the GUI to Smalltalk: https://news.ycombinator.com/item?id=49857855

I had no idea about the Interface Builder Lisp history! Thank you for digging that up!

In many ways NeXTSTEP can be thought of as a “pragmatic Smalltalk environment” that traded some of Smalltalk’s dynamism for improved compatibility with the Unix ecosystem. Recall that NeXT’s original target market was higher education, which even in the 1980s started relying on Unix. This was the era of the Unix workstation, and NeXT can be considered a competitor to Sun, SGI, HP, DEC, and other Unix workstation vendors. NeXT’s differentiator was its Smalltalk-inspired object-oriented environment, built on Objective-C.

I wonder, though, what a pure Smalltalk workstation would’ve looked like, and what hardware would’ve been necessary to make it fast. After all, the late 80s were the era of Lisp machines, though Lisp machines were much more expensive than NeXT cubes ever were. If I recall correctly, Smalltalk on the Xerox Alto was slow, but David Patterson (of RISC fame) had a PhD student who worked on a performant version of the Smalltalk VM, and some of that work went to a spiritual successor named Self, which heavily influenced the design of the Java VM. I remember Java feeling slow in the mid-1990s, but that might be due to bad memories of Java applets rather than a fundamental problem with Java VMs.

Tektronix sold Smalltalk workstations [1], you could run Lisp on them as well. You may have been able to run Smalltalk-80 on a Xerox Daybreak which was a fair bit faster than an Alto.

The first JIT was written by L Peter Deutsch for Smalltalk targeting the 68020.

[1] https://www.mirrorservice.org/sites/www.bitsavers.org/pdf/te... [2] https://en.wikipedia.org/wiki/Xerox_Daybreak

I've been looking for one of these for years. Dont think ive ever seen one for sale. Sadly my retrocimputing days (especially for such rare and likely expensive machines) is about over. I've met people willing to strike me good deals on lisp machines, but these are just a bit too obscure perhaps.

I had a brief play on a 4404 at an exhibition when they were new but have never seen any for sale since.

Yeah...wouldn't we all like to find a Tek 4404 :-)? You might like to live vicariously through this guy: https://medium.com/@adambillyard/restoring-an-ai-workstation...

He seems to have added emulation of the 4404 to Mame.

https://wirfs-brock.com/allen/things/smalltalk-things/tektro...

Your original Smalltalk port described in "Bits of History, Words of Advice" wasn't on top of an operating system but the 4404 does have one. What was the reason for this, was it to be able to run the other languages on the 4404?

Interface Builder was originally a Lisp project

https://www.inria.fr/en/jean-marie-hullot-visionary-computer...

An another lot of people doesn't know that Smalltalk-76 onwards implemented the Simula 67 object model (as the first language in the given completeness), which represented a fundamental departure from Smalltalk-72. See the Ingalls 1978 and 2020 papers for more information (https://dl.acm.org/doi/10.1145/512760.512762 and https://dl.acm.org/doi/10.1145/3386335, both more interesting and relevant than the one in the title, written by the person who invented most of the technology we today know as Smalltalk-80).

This is correct.

I got to meet Brad Cox at college when he came to visit one of my professors (who I worked with after hours). I find it interesting that instead of the objects as tiny computers of Kay he thought of software ICs and then pulled from Smalltalk to implement those concepts in Objective-C, but in the simplest way possible (almost just combining the two side-by-side) as opposed to Stroustrup being more heavily influenced by Similar (especially as he was actually building simulations) and extending C into objects to gain performance that Simula didn’t have.

s/Similar/Simula/

Similar would be the perfect name for a Simula Simlator.

Slimula is to Simula as Minyma is to Macsyma.

People naturally get hung up on the syntax and the interop story with existing popular languages, but it’s the object model differences between Smalltalk, Objective-C, Self, C++… that really matter.

ref: https://www.abebooks.com/9780201834543/Object-Model-Lippman-...

Yes, for example, while Java was inspired by C++ for its syntax and target audience, the object model and JVM semantics are closer to Smalltalk/Objective-C than C++.

The interface/@protocol, JAR files/Bundles, services/plugins, proxy objects/categories,....

When I was in college, Brad Cox came and gave a talk about Objective-C (and more generally his 'software IC' ideas). He was pretty clear how Smalltalk was a key influence. His book on ObjC discusses it as well. Very interesting stuff, even now.

Reply on news.ycombinator.com