[kabdib]

Don't know if it's true, but it makes a fun story: ST80 used reference counting with counts saturating at some low number (like 127), and of course there are problems collecting circular references. A comprehensive analysis of a modern image, with Xerox PARC lineage, discovered garbage that had been uncollected for 20-30 years.

[Rochus]

It is true, that there is uncollected garbage in the original Xerox ST80 image. I've built some tools to analyze the image and also a VM which can be interrupted at any time to analyze the current state of the image (see https://github.com/rochus-keller/Smalltalk).

There are two zombie processes (OID 6662 and 19ba). There are also a couple of BlockContext and MethodContext which have a nil sender and a reference to an unknown method, but which are still referenced from somewhere (i.e. the collection is prevented, even with mark & sweep). E.g. OID 79a2 of class BinaryChoice. I have a full list if anybody is interested.

[floren]

It's things like this that always made me look a bit askance at Smalltalk. It feels a little like "Document_Final_FINAL_v3_(2).docx", a little like a dirty old whiteboard where you can still see outlines of notes from last year. Might not be a fair assessment, but as an outsider I've always felt this way about image-based systems.

[Rochus]

Deploying an application was always an adventure. You could never be completely sure that it didn't contain things it shouldn't.

[igouy]

( cont. from https://news.ycombinator.com/item?id=31999774 )

> … no algorithm that can identify all dead code reliably…

Why were you so concerned with the removal of all dead code?

Compared to "… nobody had the source code…" it seems like a minor issue.

[igouy]

( cont. )

imo the more usual concern would be mistakenly removing code that was not dead.

Without a design document, we might think there would be no senders of #factorial without understanding that the intention was to invoke that method on the command line.

For example,

    $ cat fact.st
    Stdio stdout 
        nextPutAll: 100 factorial printString; 
        nextPut: Character lf.!
    SmalltalkImage current snapshot: false andQuit: true!

    $ bin/pharo --headless Pharo10-SNAPSHOT-64bit-502addc.image fact.st
    93326215443944152681699238856266700490715968264381621468592963895217599993229915608941463976156518286253697920827223758251185210916864000000000000000000000000

So instrumenting methods and logging — https://stackoverflow.com/a/162719

[igouy]

For some meaning of "contain things it shouldn't":

Given that the base image we started working with didn't "contain things it shouldn't" why could we not be completely sure that base image + source code fileIn didn't "contain things it shouldn't" ?

[Rochus]

First, as most other companies using ST we were not working on the original image, but on a company-specific one that had been in use for a long time and contained a lot of company-specific stuff. On the other hand, there were also things in the original images of the commercial STs that one would or should not ship with the product.

[igouy]

Firstly:

company-specific base image = base image + company-specific source code fileIn

Why could we not be completely sure that company-specific base image + source code fileIn didn't "contain things it shouldn't" ?

Secondly:

> … also things in the original images of the commercial STs that one would or should not ship with the product.

That doesn't seem to be an example of "You could never be completely sure that it didn't contain things it shouldn't."

That seems to be an example of you being completely sure.

[pjmlp]

Which is why image trimming tools did exist.

[igouy]

Is there uncollected garbage in the Smalltalk Squeak 6.0 image?

[Rochus]

I have no information about this. Squeak uses a different image format, incompatible with my tools.

[erik_seaberg]

I’m reminded of junk DNA.

[isaacimagine]

I mean, Kay was a biologist :)

[andsoitis]

Haha.

Alan Kay wrote: "At Utah sometime after Nov 66 when, influenced by Sketchpad, Simula,

the design for the ARPAnet, the Burroughs B5000, and my background in

Biology and Mathematics, I thought of an architecture for

programming. It was probably in 1967 when someone asked me what I was

doing, and I said: "It's object-oriented programming".

The original conception of it had the following parts.

  - I thought of objects being like biological cells and/or individual 

computers on a network, only able to communicate with messages (so

messaging came at the very beginning -- it took a while to see how to

do messaging in a programming language efficiently enough to be

useful)."

http://userpage.fu-berlin.de/~ram/pub/pub_jf47ht81Ht/doc_kay...

[Rochus]

It's interesting to note that he explicitly mentions "November 1966" and also refers to Simula. The ACM paper about Simula I appeared in September 1966 (see https://dl.acm.org/doi/10.1145/365813.365819), so he might have read it. Simula (or Simula I as it is called today) was a language to support discrete event simulation and had active objects (called "processes") for that purpose which "sent events" to each other (i.e. the process is activated at the event time). The language was already presented in 1962 at the second international conference on information processing held in Munich. The actually "object-oriented" Simula version with classes, inheritance and virtual methods appeared in 1967. Kay's vision applies to Smalltalk-72 to some degree, but Smalltalk-76 then also switched to inheritance and compiled virtual methods.