Dijkstra on Ada

https://craftofcoding.wordpress.com/2014/04/16/dijkstra-on-ada/

by cpeterso • 1 year ago

65 47 1 year ago

craftofcoding.wordpress.com

ggm• 1 year ago

I have it in my mind that Wirth floated modula/modula-2 variants into the early round of Ada candidates but I can find nothing evident.

My back reasoning to it's truth is threefold:

1) He did a residency at York university relating to pascal and modula in teaching CS. That's why my first uni language was pascal. (He'd just left)

2) York used pascal and modula heavily across the Ada specification window

3) York got an SERC or other funding contract to implement a multi pass Ada compiler on BSD Unix.

Which would mean (if true) Dijkstra's comments basically slated all of the candidate languages, and by implication Wirths language views, given he'd worked on the IFIP programming language specification process and was so strongly associated with pascal variants of imperative programming languages.

adrian_b2• 1 year ago

Modula was first described in March 1976.

Modula-2 is a much later language, from 1982, 3 years after the publication of Ada, and this is the version of Modula that has become well known. The first version of Modula has seen very little use, if any.

The first version of Modula has been conceived by Wirth while he took a sabbatical year during which he stayed at Xerox PARC. There he was impressed by the language Mesa, which already had modules.

In my opinion, Mesa was a programming language vastly better than any Modula version. Wirth however thought that Mesa was a too big language, so he attempted to design a simple language that offered some of the benefits of Mesa.

In 1976, when Wirth has designed the first (not very good) version of Modula, the DoD requirements specification process that has resulted in Ada had already passed through 3 stages: STRAWMAN (1975-04), WOODENMAN (1975-08) and TINMAN (1976-01), though the first requirements document that was close to the final language, IRONMAN, would be published only next year, in 1977. (The final DoD requirements document was STEELMAN, in 1978-06, and Ada was published one year later, in 1979-06.)

Therefore Modula had no influence on Ada.

After implementing the detailed DoD requirements (which included some influences from the language Jovial, which was used within the DoD at that time), the remainder of Ada was influenced by IBM PL/I, by Algol 68, by Pascal from Wirth, and also by Xerox Mesa, the source of inspiration for Modula. Any resemblance between Ada and Modula comes either from Pascal or from Mesa, which have been sources of inspiration for both Ada and Modula.

Xerox Mesa has been a very innovative programming language, but it is seldom mentioned today, despite being the origin of many features added to later languages, e.g. the less usual loop syntax used by Python (with ELSE; Mesa used a better keyword, FINISHED).

Unfortunately some of the best features of Mesa are not supported at all or only badly supported in most modern languages.

foul• 1 year ago

Unfortunately some of the best features of Mesa are not supported at all or only badly supported in most modern languages.

Excluding its epic exception handling system, what would you suggest?

pjmlp3• 1 year ago

And the foundation of one of the very first memory safe systems programming languages, with automatic resource management, using a mix of reference counting with a cycle collector, Cedar.

"Eric Bier Demonstrates Cedar"

https://www.youtube.com/watch?v=z_dt7NG38V4

Mesa also provided the very first IDE experience for strongly typed languages, with code completion, typo correction, dynamic code loading, a REPL, incremental compilation,...

"The Mesa programming environment"

https://dl.acm.org/doi/10.1145/800225.806843

Indeed, so many still catching up to Xerox PARC vision.

microtherion• 1 year ago

Counterpoint: Wirth burned out of the Algol-68 standardization effort. He definitely did not want to get into a militarized version of that process.

zombot• 1 year ago

I can't make heads or tails of this.

erk__• 1 year ago

A better article that is actually about the competition is this one which was on HN the other day: https://iment.com/maida/computer/redref/

https://news.ycombinator.com/item?id=44048775

johnisgood3• 1 year ago

So... he said this over 30 years ago? Makes Ada look bad for no reason at all, then.

f1shy• 1 year ago

Does somebody have more context about it? What were the other colors? This says absolutely nothing to me.

amszmidt• 1 year ago

What were the other colors?

ADA was designed around a competition between four companies, to make it unbiased the proposals got named by colour.

Honeywell was Green, Intermetrics was Red, SofTech was Blue, and SRI was Yellow. There where no other colors.

MisterTea2• 1 year ago

Ada is a name, not an acronym like ADA, the Americans with Disability Act, which unfortunately domain collides when searching for Ada material.

MisterTea2• 1 year ago

This was posted the other day: https://iment.com/maida/computer/redref/index.htm

phtrivier• 1 year ago

I wish the author had provided a link to the full reviews - I suspect they were more substantial.

As a aside : let's thank the FSM that Dijkstra never had access to social media - I suspect he had the kind of "abrasive" personality that would have made him probe to wasting his time and intellect arguing with all the randos of the world.

blablablerg• 1 year ago

And you don't think he would adjust like Linus for example?

enriquto• 1 year ago

maybe. But that would be so sad! (as it is with Linus)

badpenny• 1 year ago

https://www.cs.utexas.edu/~EWD/transcriptions/EWD06xx/EWD659...

https://www.cs.utexas.edu/~EWD/transcriptions/EWD06xx/EWD660...

https://www.cs.utexas.edu/~EWD/transcriptions/EWD06xx/EWD661...

https://www.cs.utexas.edu/~EWD/transcriptions/EWD06xx/EWD662...

Joker_vD2• 1 year ago

"I don't know how many of you have ever met Dijkstra, but you probably know that arrogance in computer science is measured in nano-Dijkstras." — Alan Kay

Just a quote from a review of the GREEN proposal:

They first explain the introduction of new types with "type weight = integer; type length = integer" which "define the new types "weight" and "length" as different types, both distinct from the predefined type "integer" although they have the properties of "integer"." That is a very obscure sentence: if the new types have all the properties of the old type, how can they differ?

Why, in the same way that two equal (sorry, congruent) triangles are still two different triangles. Hopefully this is not a novel concept for someone with background in maths and theoretical physics?

And this attitude of his where you can't quite pinpoint if he honestly couldn't understand something actually quite plain and simple, or just pretended not to and nitpicks, runs through almost all of his writing. The effect is that he either comes off as dim while pretentious, or just as an asshole.

And his remarks vis. statement terminators vs. separators, oh god. First of all, we sort of do have explicit statement initiators: the "function/block body begins here" token plus separators/terminators are exactly this, if you look at them from the right angle. Second, separators are just annoying for mechanistic code changes: if you switch two statements' places, but one of the was the last one in the block, you will need to also shuffle that pesky semicolon; and code generation also harder since you need to explicitly bother to not emit the separator after the last statement. And if Dijkstra has no problem with that, well, most other programmers are not him, and they do have problem with that. Not that he ever had much respect for tastes and opinions that differed from his own.

nickdothutton2• 1 year ago

Re: weight/length, he did not think you might want to define a range?

_wire_• 1 year ago

if you switch two statements' places, but one of the was the last one in the block, you will need to also shuffle that pesky semicolon...

  Blarg, zarg, arg.

  Arg, blarg. zarg,

All I want to do is switch two words places. But the pesky cap, commas and period have to be rearranged? How pretentious!

...you will need to also shuffle that pesky semicolon; and code generation also harder since you need to explicitly bother to not emit the separator after the last statement

A search for "Dijkstra on statement separators" returned a paper on programming languages and execution determinacy for mapping of concurrent and sequential programs

https://www.cs.utexas.edu/~EWD/transcriptions/EWD04xx/EWD472...

Based on this research, it's clear that Dijkstra was not troubled by the pesky syntax of statement delimiters during code translation as he was whether you would get the same program function after translation.

The papers conclusion spells this out:

//Having worked mainly with hardly self-checking hardware, with which non-reproducing behaviour of user programs is a very strong indication of a machine malfunctioning, I had to overcome a considerable mental resistance, before I found myself willing to consider non-deterministic programs seriously. It is, however, fair to say that I could never have discovered the calculus before having taken that hurdle: the simplicity and elegance of the above would have been destroyed by requiring the derivation of deterministic programs only. Whether non-determinacy is eventually removed mechanically —in order not too mislead the maintenance engineer— or (perhaps only partly) by the programmer himself because, at second thought, he does care —e.g. for reasons of efficiency— which alternative is chosen, is something I leave entirely to the circumstances. In any case we can appreciate the non-deterministic program as a helpful steppingstone.//

—

Regarding Alan Kay's quote:

His quip about nano-Dijkstras can be read as much as a statement of respect as derision.

For example, Kay's presentations occasionally repeat a rant about the hazard of semaphores for handling concurrency due to problem of deadlock. Dijkstra is known for his pioneering research on semaphores as synchronization primitives. Kay's point is well taken in a very limited context of struggles with early concurrent programming. But in the 21st century, Kay's complaint is like the Wirth's historical complaint about goto. It's true, but who thinks like that anymore?

And there's a widely held precept the abstraction is key to understanding and overcoming these complaints: you should not try to persevere writing grand plans using very low-level abstraction when you can otherwise build higher level, more appropriate abstractions from such primitives. Just as branch instructions comprise arbitrarily higher orders program logic, so do semaphores comprise higher order constructs for synchronization.

Everything about Alan Kay shows he very well understands abstraction, and he studied to be a mathematician, so Dijkstra's formalism is familiar to him. Well, a great jazz riff always has some off-notes.

So prefer to read the Kay quote as pithy, not derisory.

Surac• 1 year ago

That settles it for now and forever. Men are only able to distinguish 6 colors.

rini17• 1 year ago

The actual modern Ada ended up being reasonable and the conservative design is justified. And it's no more verbose than Java anyway. Is there any writeup on the discrepance between the hysteria and the real thing?

I think it was blowback to the hype. Rust, sadly, seem to be bound to end likewise as the rustaceans enthusiastically endorse AI-driven rewriting of legacy DoD C code into Rust. What can go wrong there?

Ygg22• 1 year ago

And it's no more verbose than Java anyway.

That's damning with faint praise. Java is one of the most verbose languages.

Rust, sadly, seem to be bound to end likewise as the rustaceans enthusiastically endorse AI-driven rewriting of legacy DoD C code into Rust.

Sure, some people want to speed up and/or automate conversion C to Rust, they are a minority. Going for 1 to 1 translation is going to be problem fraught with issues.

pjmlp3• 1 year ago

I love verbosity more than programming with hieroglyph languages as so common in UNIX culture, need to save those keyboards longevity, a side effect from subpar IDE tooling.

Microsoft is actually doing this as well, as part of their in-house migration to Rust, see RustNation UK 2025 talks from Microsoft.

jvanderbot2• 1 year ago

As a rustacean I am appalled by the idea of AI driven rewriting of legacy C. Is this a real thing??

Ygg22• 1 year ago

You don't know about TRACTOR?

https://www.darpa.mil/program/translating-all-c-to-rust

hello_computer2• 1 year ago

Did Dijkstra have anything nice to say about anything?

Rochus2• 1 year ago

Dijkstra showed particular appreciation for Haskell (see http://www.cs.utexas.edu/users/EWD/OtherDocs/To%20the%20Budg...).

hello_computer2• 1 year ago

His magician archetype was out-of-control, but at least he had taste.

toolslive2• 1 year ago

Haskell, though not perfect, is of a quality that is several orders of magnitude higher than Java.

I'm pretty sure he would assess Python a few orders of magnitude lower than Java. Probably warranting the need for a log scale for the quality of programming languages.

bazoom423• 1 year ago

Hilariously he begins the petition by disparaging the university, the city, and the entire state. I wonder how well that went?

mcv• 1 year ago

The guy certainly had a brutal way with words.

cess112• 1 year ago

One could also describe him as direct, concise and unambigous except as a joke.

bazoom423• 1 year ago

There is no technical substance in the quotes though. You could use them against whatever language you don’t like. I’m sure they are taken out of context, but I find it weird how people are so impressed by these generic insults.

cyco130• 1 year ago

The green proposal was the one that got picked up to become Ada by the way. Red was the runner-up.

johnisgood3• 1 year ago

I need more context, I do not understand anything apart from the quote on the bottom, which is just a mere expression of dislike of something without any reasons provided...

Ada was such a mess that I shuddered at the thought that Western security would depend on it and that I would feel much safer if the Red Army were to adopt it as well.

Replace "Ada" with "Rust" or anything else (like Dijkstra's favorite programming language). It does not explain the why.

nickdothutton2• 1 year ago

Did not disappoint. There is a reason we have a term "Dutch uncle". RIP EWD.

Dijkstra on Ada

Related Stories