On 09/09/06, David Leangen <[EMAIL PROTECTED]> wrote:
My pleasure.
Correct.
Intuitive reasoning certainly makes fewer demands. When you start to formalise something, you realise pretty quickly how much you really don't understand very well. But that sometimes beats finding out later ( e.g. during testing!). For example, early on in my work on OSGi R4, I wrote down some formal definitions relating to class loading and then used those to check my informal thinking from time to time. Unfortunately however such formal definitions aren't a great communication mechanism since not many software developers have the necessary maths background although quite a few computer science degrees include a formal specification module so developers sometimes can read a spec. even if they couldn't really write one.
Glyn
Thanks, Glyn,
I forgot about the mathematical sense of the word. Thanks for the nice
reminder!
My pleasure.
Ok, so, if want to apply mathematical reasoning to the dependency problem
(and my math is VERY rusty...), it seems that the set of declared
dependencies would be irreflexive, since we do not declare a bundle as a
dependency of itself.
Correct.
In terms of closure, it would be nice to find some kind of formalism for a
closed set, but if x R y and y R z, the set is only closed if z has no other
dependencies than x and y, so introducing closure into the equation seems
somewhat complicated in practice. Seems that, as you pointed out, at this
point intuitive reasoning seems easier than formal reasoning...
Intuitive reasoning certainly makes fewer demands. When you start to formalise something, you realise pretty quickly how much you really don't understand very well. But that sometimes beats finding out later ( e.g. during testing!). For example, early on in my work on OSGi R4, I wrote down some formal definitions relating to class loading and then used those to check my informal thinking from time to time. Unfortunately however such formal definitions aren't a great communication mechanism since not many software developers have the necessary maths background although quite a few computer science degrees include a formal specification module so developers sometimes can read a spec. even if they couldn't really write one.
Glyn
> -----Original Message-----
> From: [EMAIL PROTECTED]
> [mailto:[EMAIL PROTECTED]]On Behalf Of Niclas Hedhman
> Sent: 8 September 2006 20:34
> To: General OPS4J
> Subject: Re: Transitive dependency
>
>
> On Friday 08 September 2006 18:39, Glyn Normington wrote:
> > I'll refrain from expanding the formal mathematical definition as the
> > intuitive notion is more important in computing.
> >
> > A familiar example could be an 'impactedBy' relation defined as the
> > reflexive, transitive closure of the 'dependsOn' relation since
> a bundle,
> > for example, is obviously impacted by changing the things it depends on
> > (transitively ;-) ) as well as by changing the bundle itself.
>
> Ok, I get it...
>
> Thanks a lot!
> Niclas Hedhman
>
> _______________________________________________
> general mailing list
> [email protected]
> http://lists.ops4j.org/mailman/listinfo/general
>
_______________________________________________
general mailing list
[email protected]
http://lists.ops4j.org/mailman/listinfo/general
_______________________________________________ general mailing list [email protected] http://lists.ops4j.org/mailman/listinfo/general
