3 ms·

The two hardest problems are naming, cache invalidation, and off-by-one errors. In this case, there's a naming problem. "Class 1" and "Class 2" don't give any

by dsr_ 5y ago

The two hardest problems are naming, cache invalidation, and off-by-one errors. In this case, there's a naming problem. "Class 1" and "Class 2" don't give any semantic content. Is there a "Class 3" coming along? Let me suggest "Inadequacy Problems" and "Ubiquity Problems" as alternatives.

webmaven 5y ago
> The two hardest problems are naming, cache invalidation, and off-by-one errors. > In this case, there's a naming problem. "Class 1" and "Class 2" don't give any semantic content. To me, the names suggested a connection or analogy to Type 1 and Type 2 errors, also known as False Positives and False Negatives. I was wrong, as it turns out, but the opaqueness of the labels, while annoying and unnecessary, isn't unprecedented.
sudhirj 5y ago
> Is there a "Class 3" coming along? Quite possibly. There is a class of problems to deal with when technology has outlived its usefulness and must be gotten rid of. Coal power arguably has Class 3 problems now. There are also Class 0 problems, like fusion, where just getting things off the ground is difficult.
- dsr_ 5y ago
  "[Dis-]Graceful termination" and "Theory-to-practice" are good names for those problems. They supply useful context and are not ambiguous. Class 1: medical devices, explosives, renewable energy credits, depreciable property, lasers, streetcars, UK national insurance, railroads, electronics, felonies... If you're going to coin a term, make it a good one.
dragontamer 5y ago
> The two hardest problems are naming, cache invalidation, and off-by-one errors. Aside: the joke I've heard was always naming, resource management, and off-by-one errors. Resource management would include cache invalidation, malloc/free, open()/close(), accept()/close(), dup()/close(), garbage collection, etc. etc.
- qsort 5y ago
  I'm familiar with your parent's variant ("cache invalidation") as well. https://skeptics.stackexchange.com/questions/19836/has-phil-karlton-ever-said-there-are-only-two-hard-things-in-computer-science#39178 https://skeptics.stackexchange.com/questions/19836/has-phil-...
- olau 5y ago
  If we're going to generalize this old saying, let me point out that naming is just a specific instance of resource management, name space management. So I guess from this perspective, there's really only one hard problem in computer science: resource management and off-by-one errors.

by dsr_ 5y ago

webmaven 5y ago

> The two hardest problems are naming, cache invalidation, and off-by-one errors. > In this case, there's a naming problem. "Class 1" and "Class 2" don't give any semantic content. To me, the names suggested a connection or analogy to Type 1 and Type 2 errors, also known as False Positives and False Negatives. I was wrong, as it turns out, but the opaqueness of the labels, while annoying and unnecessary, isn't unprecedented.

sudhirj 5y ago

> Is there a "Class 3" coming along? Quite possibly. There is a class of problems to deal with when technology has outlived its usefulness and must be gotten rid of. Coal power arguably has Class 3 problems now. There are also Class 0 problems, like fusion, where just getting things off the ground is difficult.

dsr_ 5y ago
"[Dis-]Graceful termination" and "Theory-to-practice" are good names for those problems. They supply useful context and are not ambiguous. Class 1: medical devices, explosives, renewable energy credits, depreciable property, lasers, streetcars, UK national insurance, railroads, electronics, felonies... If you're going to coin a term, make it a good one.

dragontamer 5y ago

> The two hardest problems are naming, cache invalidation, and off-by-one errors. Aside: the joke I've heard was always naming, resource management, and off-by-one errors. Resource management would include cache invalidation, malloc/free, open()/close(), accept()/close(), dup()/close(), garbage collection, etc. etc.

qsort 5y ago
I'm familiar with your parent's variant ("cache invalidation") as well. https://skeptics.stackexchange.com/questions/19836/has-phil-karlton-ever-said-there-are-only-two-hard-things-in-computer-science#39178 https://skeptics.stackexchange.com/questions/19836/has-phil-...
olau 5y ago
If we're going to generalize this old saying, let me point out that naming is just a specific instance of resource management, name space management. So I guess from this perspective, there's really only one hard problem in computer science: resource management and off-by-one errors.