1 |
2011/6/23 Duncan <1i5t5.duncan@×××.net>: |
2 |
> Jesús J. Guerrero Botella posted on Thu, 23 Jun 2011 08:15:44 +0200 as |
3 |
> excerpted: |
4 |
> |
5 |
>> Symlinks are clean, and portage has always been file-oriented so I see |
6 |
>> no problem with using them for this. |
7 |
> |
8 |
> It has been some years since I've seen the argument made, but if I'm not |
9 |
> mistaken, at least back in 2004-ish when I first switched to Gentoo, the |
10 |
> argument against in-tree symlinking (or multi-hard-linking, for that |
11 |
> matter) of any kind (other than the obvious directory hard-linking) was |
12 |
> that we wanted to keep the tree at least minimally deployable on non-Unix |
13 |
> filesystems like fat/ntfs. Note that while a user's profile uses a |
14 |
> symlink, the symlink is on /etc (which is thus implied to be a Unix |
15 |
> filesystem with symlinking capacities) pointing /into/ the tree, NOT |
16 |
> actually PART OF the tree. |
17 |
> |
18 |
> One scenario in which this might be a factor is that of someone doing |
19 |
> their syncs and source downloads at work where they have lots of |
20 |
> bandwidth available, then sneakernetting it home on a fat32 formatted |
21 |
> thumbdrive. |
22 |
> |
23 |
> Now it can be argued that the flexibility benefit of multi-category |
24 |
> packages trumps that of being able to put the tree on fat or whatever, |
25 |
> but there IS a definite loss of tree portability that's implied, and thus |
26 |
> a tradeoff to be considered. |
27 |
|
28 |
Yes, that's true. But it's also true that besides the symlinks, the |
29 |
portage tree will be broken the same moment you put it into a fat |
30 |
volume, because it will directly erase all the permissions and |
31 |
ownership metadata. So, the thing is already broken, why should we |
32 |
care if it breaks a bit more? Seriously, limiting ourselves because of |
33 |
an fs that not even MS uses any longer is not a smart thing to do, in |
34 |
my opinion. |
35 |
|
36 |
-- |
37 |
Jesús Guerrero Botella |