| 1 |
On Sat, 31 Mar 2012 18:39:21 +0300 |
| 2 |
Alex Alexander <alex.alexander@×××××.com> wrote: |
| 3 |
> On Mar 31, 2012 5:57 PM, "Ciaran McCreesh" |
| 4 |
> <ciaran.mccreesh@××××××××××.com> wrote: |
| 5 |
> > On Sat, 31 Mar 2012 15:08:29 +0300 |
| 6 |
> > Alex Alexander <alex.alexander@×××××.com> wrote: |
| 7 |
> > > No. I didn't say I think it works, I said I have proof it works. |
| 8 |
> > |
| 9 |
> > Well that's interesting, because there are plenty of examples where |
| 10 |
> > it doesn't work, and all that it takes to disprove a theory is a |
| 11 |
> > single counterexample. So I think you're misunderstanding what |
| 12 |
> > constitutes proof here -- "some evidence" certainly isn't it. |
| 13 |
> |
| 14 |
> Boring. You conveniently ignored the other part of my message. |
| 15 |
|
| 16 |
I was hoping you'd understand how your claim of proof was utterly |
| 17 |
unfounded. What you have is limited evidence of a very specific |
| 18 |
situation, which is a whole other thing. |
| 19 |
|
| 20 |
> I'll repeat it: no matter how much you argue, it'll still work fine |
| 21 |
> for me. |
| 22 |
|
| 23 |
And I'll spell it out. On the scale of "it works", you have a series of |
| 24 |
levels: |
| 25 |
|
| 26 |
1. It compiles. |
| 27 |
|
| 28 |
2. I tried it and I didn't see any breakages. |
| 29 |
|
| 30 |
3. I tried it, and I checked carefully that nothing was broken. |
| 31 |
|
| 32 |
4. I tried it on a wide range of valid inputs, and I checked carefully |
| 33 |
that nothing was broken. |
| 34 |
|
| 35 |
5. I tried it on a wide range of valid inputs, including inputs designed |
| 36 |
to test edge cases, and I checked carefully that nothing was broken. |
| 37 |
|
| 38 |
6. I tried it on a wide range of valid and invalid inputs, and I |
| 39 |
checked carefully that nothing was broken, and that the invalid inputs |
| 40 |
were handled correctly. |
| 41 |
|
| 42 |
7. I carefully considered all the equivalence classes of inputs, and |
| 43 |
tested each. |
| 44 |
|
| 45 |
8. I carefully considered all the equivalence classes of inputs, and |
| 46 |
can explain why each is handled correctly. |
| 47 |
|
| 48 |
9. I can prove that it works. |
| 49 |
|
| 50 |
You're offering evidence of number 2, or possibly 3. Gentoo is a large |
| 51 |
system containing many interacting components, that is expected to keep |
| 52 |
working for long periods in many different unpleasant situations. We |
| 53 |
need to be at at least number 7 here, and ideally number 8. For preserve |
| 54 |
libs, the feature fails at number 4. |
| 55 |
|
| 56 |
-- |
| 57 |
Ciaran McCreesh |
Archives