1 |
On Monday 16 August 2010 11:38:17 Dale wrote: |
2 |
> Peter Humphrey wrote: |
3 |
> > On Monday 16 August 2010 09:13:29 Dale wrote: |
4 |
> >> Massimiliano Ziccardi wrote: |
5 |
> >>> # equery belongs /usr/lib/libxfce4util.la<http://libxfce4util.la> |
6 |
> >>> [ Searching for file(s) /usr/lib/libxfce4util.la |
7 |
> >>> <http://libxfce4util.la> in *... ] |
8 |
> >>> # |
9 |
> >>> " |
10 |
> >> |
11 |
> >> Equery doesn't give any results because it is not installed. |
12 |
> > |
13 |
> > If it weren't installed it wouldn't be able to announce what it was |
14 |
> > searching for and where :-) |
15 |
> |
16 |
> Some new people think it knows where things come from even if it is not |
17 |
> installed. It can't do that so I posted that in case the person didn't |
18 |
> know. Then I posted a way to find out even if a package is not |
19 |
> installed. I didn't know about that website until someone pointed it |
20 |
> out to me many ages ago. |
21 |
> |
22 |
> It would be neat if it could do that tho. Just have no idea how it |
23 |
> could. ;-) |
24 |
|
25 |
It would have to consult a database of file->package mappings. For already |
26 |
installed packages, it simply looks in /var/ locally - an excellent high- |
27 |
quality database of such already exists there :-) |
28 |
|
29 |
For not already installed packages, it would have to look out on the |
30 |
intartubes for the site you mentioned or something similar. A portage dev |
31 |
would have to be batshit insane to even try that as the results are "fuzzy". |
32 |
There are filename collisions to deal with, plus the fact that you don't know |
33 |
what a package installs till someone builds it. And then there's the carnage |
34 |
caused by USE flags - my list of installed files for a package is likely |
35 |
different to yours. |
36 |
|
37 |
A hallmark of portage is that it deals in definitive data - results are always |
38 |
exact, unambiguous and deterministic - especially deterministic. |
39 |
|
40 |
Google for instance, is none of these things. Google results are "good |
41 |
enough". Mixing portage's view of data with google's view of data will break |
42 |
portage is many many horrible ways and make your life (as someone who helps |
43 |
others on this list) miserable - users *will* trust such a lookup tool to be |
44 |
definitive. |
45 |
|
46 |
Which all goes to say you should resist the creation of any such lookup tool |
47 |
with every fibre of your being :-) |
48 |
|
49 |
|
50 |
-- |
51 |
alan dot mckinnon at gmail dot com |