1 |
On Fri, 11 Nov 2005 18:40:53 +0100 Marius Mauch <genone@g.o> |
2 |
wrote: |
3 |
| I've already asked a similar question in another mail (in other |
4 |
| context) without an answer, but how many news items do people believe |
5 |
| will exist at any given time? Depending on the actual implementation |
6 |
| it might be required to scan all existing news items which could take |
7 |
| some time if there is a large number of them (say, more than a few |
8 |
| hundred) It's clear that noone can present accurate numbers, just |
9 |
| after some rough estimates here. |
10 |
|
11 |
I don't expect it to be of the order of a few hundred. AFAIK there |
12 |
aren't huge numbers of nasty upgrades... |
13 |
|
14 |
| Also it might be useful for this filtering to be an external tool |
15 |
| using portage functions and called by portage (see also below). |
16 |
| Although this could be considered an implementation detail as it's |
17 |
| mostly transparent for ebuild devs/users. |
18 |
|
19 |
Yeah, I have a script which does it already. It's just rather slow |
20 |
because of portageq. It'll be in the next draft. |
21 |
|
22 |
| - local storage of news items / "read" attribute: |
23 |
| I don't really the like the copy-if-new feature and handling at the |
24 |
| fs level, I'd use a file that lists the ids of news items (potentially |
25 |
| with a timestamp and/or status field). I don't see any benefits of |
26 |
| having redundancy here, and accessing the news in $PORTDIR is simple |
27 |
| enough. Granted race conditions might be an issue (where the solution |
28 |
| complicates tools), but that's so minor I wouldn't really care about |
29 |
| it. |
30 |
|
31 |
I'll have to think about that one... ID lists should be easy from an |
32 |
implementation perspective... |
33 |
|
34 |
| Another thing that's unclear: "Whenever relevant unread news items are |
35 |
| found, emerge will copy or symlink the news file into |
36 |
| /var/lib/gentoo/news/." |
37 |
| This "Whenever ... are found" is too vague, should this apply to |
38 |
| emerge --sync, any emerge operation, any "import portage" call or |
39 |
| what? This isn't just an implementation detail. |
40 |
|
41 |
I'd say after emerge --sync, plus after an emerge --pretend and before |
42 |
an emerge blah. Will there be hooks for these? |
43 |
|
44 |
| PS: I'd avoid symlinks here at all costs, too easy to break them. |
45 |
|
46 |
Probably true, especially if portdir changes... |
47 |
|
48 |
| Also as Brian and Jason have said already, the system should be able |
49 |
| to handle multiple repositories. So to use the current GLEP as |
50 |
| example, at least the path should be changed |
51 |
| to /var/lib/gentoo/news/porttree |
52 |
|
53 |
Pfff, if that ever happens it's just a case of adding in another |
54 |
directory level. As it stands though, there's no multiple repo support |
55 |
in portage and no way to uniquely identify a repo, so I see no point in |
56 |
going out of the way to handle something which may or may not ever |
57 |
happen. |
58 |
|
59 |
|
60 |
| - quality control: |
61 |
| "Any complaints regarding wording or clarity must be addressed before |
62 |
| the news item goes live." Playing devils advocate here, but complaints |
63 |
| regarding correctness are not mandatory to be adressed? |
64 |
|
65 |
Mmm, guess I should change that to "Any complaints (including, without |
66 |
prejudice to the aforesaid generality, those of wording, clarity or |
67 |
correctness) must ...". |
68 |
|
69 |
| As for using -core, please add a small explanation or at least a |
70 |
| reference to the appropriate policy docs |
71 |
|
72 |
I've moved the -core stuff to a .. Note:: for the next draft. |
73 |
|
74 |
| Things that should definitely be changed: |
75 |
| - Integration with existing systems: |
76 |
| This definitely should be a requirement for the GLEP to be considered |
77 |
| final. It doesn't prevent some thing being implemented sooner than |
78 |
| others, but multi-channel distribution (to use a buzzword) is a |
79 |
| requirement from where I come from. |
80 |
|
81 |
I'd really rather that news.gentoo.org / news2announceemail / whatever |
82 |
were handled via separate GLEPs. 42 is fairly long as it is... |
83 |
|
84 |
-- |
85 |
Ciaran McCreesh : Gentoo Developer (Look! Shiny things!) |
86 |
Mail : ciaranm at gentoo.org |
87 |
Web : http://dev.gentoo.org/~ciaranm |