Linus Says No to 'Specs'

auckland map writes to tell us about an interesting debate that is being featured on KernelTrap. Linus Torvalds raised a few eyebrows (and furrowed even more in confusion) by saying "A 'spec' is close to useless. I have _never_ seen a spec that was both big enough to be useful _and_ accurate. And I have seen _lots_ of total crap work that was based on specs. It's _the_ single worst way to write software, because it by definition means that the software was written to match theory, not reality."

Comments

Re:Linus Taken to Task

[Rakshasa Taisab]

The summary is twisting the discussion into something very flamebait worthy, and it gives a false impression of what Linus really said.

From my impression of the discussion it isn't specs he is against, but rather following those specs without being flexible enough to take reality into account. Would you really want to fly the plane built to spec, if some of those specs turned out to not accurately reflect reality? In that case you'd change the specs, but that's not always possible.

In this specific case, the problem was someone using the abstraction layering described in the spec, while the kernel would according to the others be better off using another design. If the software behaves equivalently, it should not matter how it is designed internally.

Re:Linus Taken to Task

[Lodragandraoidh]

I think there is some confusion between the idea of 'software specifications' and 'standard interface specifications'. One may referenct the other, but they are not the same.

Software specifications define functional, design and sometimes implimentation details, and in my experience never survive implimentation simply because the developer can never anticipate every obstacle that reality throws at it. Thus the software specification quickly is changed, after the fact, to represent what was implimented. It lags behind reality and serves only to document what *was* done, rather than driving the development.

Standard interface specifications (APIs, and other standards such as POSIX) serve as agreements between many different players about how a particular interface should work to provide interoperability between all users of the standard. If you want to foster interoperability then you follow the standards that are widely accepted. Without standard compliance Linux would not be where it is today.

I read the article and it occurs to me that the confusion between them was at the heart this very issue. This confusion is sadly furthered by the use of the term 'specification' to represent both things - which are clearly different. Implimentation of emergent technology is different than established protocol agreements. Perhaps we should couch our language to make those differences clear when we speak about it.