| 1 |
On Thursday 17 January 2013 14:44:14 Ciaran McCreesh wrote: |
| 2 |
> On Thu, 17 Jan 2013 14:35:12 -0500 James Cloos wrote: |
| 3 |
> > >>>>> "CM" == Ciaran McCreesh writes: |
| 4 |
> > CM> That's what's known as "doing it wrong". You should be querying |
| 5 |
> > CM> your package mangler for a list of categories, not doing an 'ls'. |
| 6 |
> > |
| 7 |
> > ls(1) isn't relevant. find(1) is. grep(1) is. There are others. |
| 8 |
> > |
| 9 |
> > Using the 'package managers' isn't very helpful. They generally do |
| 10 |
> > everything poorly. And usually **s*l*o*w*l*y**, if they compile at |
| 11 |
> > all. |
| 12 |
> |
| 13 |
> On the other hand, they do things correctly, which your approach |
| 14 |
> doesn't. |
| 15 |
> |
| 16 |
> > I can't even remember every time I've needed to use a regex, glob or |
| 17 |
> > other pattern match where the fact that the real categories had a dash |
| 18 |
> > made things easier and faster. |
| 19 |
> |
| 20 |
> But wrong. If you want wrong answers quickly, cat /dev/urandom. |
| 21 |
|
| 22 |
and breaking people for no good reason is just that -- not a good reason. |
| 23 |
|
| 24 |
is code that makes this assumption kind of crappy ? yes. is this new |
| 25 |
proposal a compelling use case for breaking that (pretty common) assumption ? |
| 26 |
no. there's no real technical overhead to have new qt categories follow the |
| 27 |
existing practice. |
| 28 |
-mike |
Archives