1 |
On 2018.11.30 17:06, Michał Górny wrote: |
2 |
> On Mon, 2018-11-26 at 21:43 +0000, Roy Bamford wrote: |
3 |
> > On 2018.11.26 18:58, Michał Górny wrote: |
4 |
> > > Here's the newest version. |
5 |
> > > |
6 |
> > > Changes: |
7 |
> > > |
8 |
> > > - added explicit notion of parent directory (missing in previous |
9 |
> GLEP |
10 |
> > > but present in implementation), |
11 |
> > > |
12 |
> > > - explicitly named GNU tar format with list of permitted |
13 |
> extensions, |
14 |
> > > |
15 |
> > > - changed volume label to 'gpkg-1.txt' file to improve |
16 |
> portability; |
17 |
> > > made |
18 |
> > > it explicit version identifier as well, |
19 |
> > > |
20 |
> > > - added info on other package formats to rationale. |
21 |
> > > |
22 |
> > |
23 |
> > [snip] |
24 |
> > |
25 |
> > The image archive stores all the files to be installed by the binary |
26 |
> > package. It should be included as the last of the files in the |
27 |
> binary |
28 |
> > package container. |
29 |
> > |
30 |
> > [snip] |
31 |
> > > |
32 |
> > > -- |
33 |
> > > Best regards, |
34 |
> > > Michał Górny |
35 |
> > > |
36 |
> > |
37 |
> > Its a nit today but that says that any future extensions, none |
38 |
> > yet planned, should be placed before the image archive. |
39 |
> |
40 |
> Yes. |
41 |
> |
42 |
> > The specification needs to avoid the use of relative references. |
43 |
> > |
44 |
> |
45 |
> I don't understand. Could you be more specific what you expect |
46 |
> instead? |
47 |
> |
48 |
> -- |
49 |
> Best regards, |
50 |
> Michał Górny |
51 |
> |
52 |
|
53 |
Michał, |
54 |
|
55 |
Enumerate the elements, in the preferred order, which you have |
56 |
already done. The is no need, in a specification that is intended |
57 |
to be easily extensible to specify that any element should be last. |
58 |
That constrains extensions. |
59 |
|
60 |
To build on an example extension given earlier. Suppose an |
61 |
extension came along to add the ebuild, required eclasses and |
62 |
sources. The present wording says that they should be included |
63 |
before image archive. |
64 |
|
65 |
Implementations may be capable of working with partial |
66 |
downloads, why force the download of elements that may not be |
67 |
required to get the payload. |
68 |
|
69 |
The overhead of the presently define elements is small compared |
70 |
to the image and its useful to be able check the metadata to |
71 |
determine if the image is really what is required. |
72 |
|
73 |
image 'last' works with the presently defined elements but may |
74 |
not be so good in the years to come. |
75 |
|
76 |
Its a subtle difference between 'last', which means always at |
77 |
the end, no mater what, and 'fifth' which is last today but |
78 |
might not be in the future. |
79 |
|
80 |
-- |
81 |
Regards, |
82 |
|
83 |
Roy Bamford |
84 |
(Neddyseagoon) a member of |
85 |
elections |
86 |
gentoo-ops |
87 |
forum-mods |