conway's law: any organization which designs a system will inevitably produce a design whose structure is a copy of the organization's communication structure.

computer scientists follow the salami slice tactic, says richard gabriel. it appears to them to be 'natural' just to accept papers that differ the least. yet "their work must continue ... thefoundation (foo/bar, 472 KB)

one can't blame tony hoare of doing that. his project the verifying compiler doesn't sound like a small slice. i am still not sure if he got all the money yet, even there are some indications that he might be on track.
tony hoare wants to dedicate 15 years of research to find a way to fix bugs. - that's the time span of grand challenge projects by definition. (one might say since the genome project, i don't know.) in short: he wants to deliver a database with code for further use: tested and deemed as correct to start with.

Premature optimization is the root of all evil in programming. C. A. R. Hoare (quoted by r.garbiel in ArtOfLisp)

according to him "software works because people designed it to work; over time hard- and software has changed, but what remains is the old (foo/bar, 684 KB) code. "we now have to look at the remaining code as if it were a natural phenomenon (foo/bar, 621 KB) .

at the event where i heard tony hoare presenting his vision was also niklas wirth, the old swiss man, who once brought schoolkids pascal. he wasn't really amused of hoare: "as an engineer" he has some problems with the idea, because it will occupy a lot of people's mind for a long time; too long in his mind. and second because he doubts its performance: "what's my advantage if there are trains that arrive 100% on time, but consist of nothing more than uncomftable boxcars?"

what't the use of a programm, that's 100% korrekt (foo/bar, 314 KB) ? and what means verifikation (foo/bar, 254 KB) if also the formal specifications can be wrong, because they are made by humans?

in certain, well-defined areas, it's maybe secure to let genetic algorithms do the job, r.gabriel suggests. anyhow, it would be more important to teach some nuances in programming as when and where it's better not to fix an error than to fix it: he gives credit - if i understand it right - to martin rinard, for the line 'forgiving and unforgiving regions of code' and adds: "we are just at the beginning (foo/bar, 803 KB) ...

b. randell recalls 1968 and one of the early conferences on software engineering. at that time "software became to have meaning in normal business parlance" yet problems sound similar: reliability, bugs, software as a commodity, modularity and structuring, problem of large systems. maybe also the following quote (found in randell's paper) is not yet outdated.

"A contract research organisation had eight people who were to produce a COBOL and an ALGOL compiler. After some initial estimates of difficulty and time, five people were assigned to the COBOL job and three to the ALGOL job. The resulting COBOL compiler ran in five phases, the ALGOL compiler ran in three." (m.e. conway, how do committees invent?, datamation, 1968).

melvin conway 40years later ...