3 ms·

This seems like a really lousy, not to mention old (2011), article. The function of Knuth’s expository pascal program is not to sort input data, much less to d

This seems like a really lousy, not to mention old (2011), article. The function of Knuth’s expository pascal program is not to sort input data, much less to demonstrate "the best" way to go about doing so. Rather it's wholly intended to show all the important details of his "literate programming" proposal. Loudly proclaiming that there's a "better" method of sorting input data using common Unix binaries and shell scripting is to proclaim loudly that you've nearly entirely missed the point of the proposal.

amouat 9y ago
Or perhaps McIllroy fully understood that and wanted to make the point that programming with higher-level abstractions makes software easier to document and understand. It's hard to say without being able to read the original article. I found it fascinating anyway. EDIT: And of course you can find the article on-line fairly easily. Some quotes from the review: "the paper eloquently attests that the discipline of simultaneously writing and describing a program pays off handsomely." "Knuth’s purpose was to illustrate WEB. Nevertheless, it is instructive to consider the program at face value as a solution to a problem. A first engineering question to ask is: how often is one likely to have to do this exact task’? Not at all often, I contend. It is plausible, though, that similar, but not identical, problems might arise. A wise engineering solution would produce-or better, exploit-reusable parts. "
- zeckalpha 9y ago
  I'd argue McIllroy's solution was literate, too, given the description.
- jfk13 9y ago
  Someone must ultimately have written those higher-level abstractions (such as tr, sort, uniq, sed, etc.) in a lower-level language. In a literate form? (Unlikely!) Of course it's generally easier to document and understand software built using (appropriate) higher-level abstractions. Just like most of us find it easier to write and maintain {C++,Java,Python,JS,Lisp,...} software than to do everything in machine code. But whatever level we're programming at, there is arguably a place for a literate programming approach to improve exposition. Indeed, I've seen *nix shell scripts that are large and complex enough that a literate programming style would be a huge benefit in attempting to understand them.
  - amouat 9y ago
    Sure. And having read (part of) the paper, the point was more about reuse than abstraction (though they are obviously related concepts).

amouat 9y ago

Or perhaps McIllroy fully understood that and wanted to make the point that programming with higher-level abstractions makes software easier to document and understand. It's hard to say without being able to read the original article. I found it fascinating anyway. EDIT: And of course you can find the article on-line fairly easily. Some quotes from the review: "the paper eloquently attests that the discipline of simultaneously writing and describing a program pays off handsomely." "Knuth’s purpose was to illustrate WEB. Nevertheless, it is instructive to consider the program at face value as a solution to a problem. A first engineering question to ask is: how often is one likely to have to do this exact task’? Not at all often, I contend. It is plausible, though, that similar, but not identical, problems might arise. A wise engineering solution would produce-or better, exploit-reusable parts. "

zeckalpha 9y ago
I'd argue McIllroy's solution was literate, too, given the description.
jfk13 9y ago
Someone must ultimately have written those higher-level abstractions (such as tr, sort, uniq, sed, etc.) in a lower-level language. In a literate form? (Unlikely!) Of course it's generally easier to document and understand software built using (appropriate) higher-level abstractions. Just like most of us find it easier to write and maintain {C++,Java,Python,JS,Lisp,...} software than to do everything in machine code. But whatever level we're programming at, there is arguably a place for a literate programming approach to improve exposition. Indeed, I've seen *nix shell scripts that are large and complex enough that a literate programming style would be a huge benefit in attempting to understand them.
- amouat 9y ago
  Sure. And having read (part of) the paper, the point was more about reuse than abstraction (though they are obviously related concepts).