1 |
Stuart Herbert kirjoitti (sunnuntai, 21. maaliskuuta 2004 21:21): |
2 |
> On Saturday 20 March 2004 12:06, John Nilsson wrote: |
3 |
> > I never ment that Gentoo developers would do some non standard things to |
4 |
> > the packages. If the standard is not good enough to convince upstream |
5 |
> > developers to change, it is not good enough, period. |
6 |
> |
7 |
> It'll never happen. XML isn't something that everyone accepts is always a |
8 |
> good thing. In many languages, handling XML content programmatically is a |
9 |
> RightRoyalPain(tm). And we currently lack adequate XML-aware diff and grep |
10 |
> tools - two things essential to making this work. |
11 |
|
12 |
This is true. One option, which would make the solution considerably iffier, |
13 |
but ease the transition from the current model quite a bit, would be to use a |
14 |
configuration file generator for each package. I.e. the configuration tool |
15 |
would read the possible options and documentation from the schema and then |
16 |
call for the config file generator to write the file according to the users |
17 |
choices. Thus the configuration file itself (and the main package) wouldn't |
18 |
have to change at all. |
19 |
Now obviously the file generator would have to be maintained upstream, |
20 |
otherwise it'd be linuxconf all over again, except worse. However, this way |
21 |
the upstream packages themselves wouldn't need any major overhauls, but just |
22 |
a compliance layer (consisiting of the schema and the converter) between it's |
23 |
native configuration and the common protocol. |
24 |
This apporach could also solve the issues of handling XML in various |
25 |
languages. And yes, I know it's a maintenance hell. |
26 |
|
27 |
> So how would the config tool know how to generate the XML file? A schema |
28 |
> is nowhere near enough. |
29 |
|
30 |
If both the schema and the configuration file would adhere to a common |
31 |
standard (which, AFAICT is what we're talking about), it shouldn't be a |
32 |
problem. |
33 |
|
34 |
> How many packages are there in the Gentoo tree? Over six thousand. Let's |
35 |
> say you could convert each one in an average of one man day's effort. I |
36 |
> think that's optimistic, but it keeps it simple. Now, a man year is around |
37 |
> 200 days of effort, allowing for weekends, holidays, illness, and |
38 |
> overheads. So that's 30 man years of effort to convert what we have today. |
39 |
|
40 |
Except that the vast majority of those packages either don't even have |
41 |
configuration files, or are niche packages, which on a large scale don't |
42 |
really matter. |
43 |
IMHO having support from just c. 50-100 major packages would be enough to |
44 |
make the protocol a true standard and with 200-300 packages it would already |
45 |
change the way all Linux systems are maintained and configured. Whether |
46 |
packages like dailystrips or xmms would follow is largely moot. |
47 |
|
48 |
> If you had RedHat's money, maybe it could be done. But Gentoo people are |
49 |
> volunteers. On average, volunteers on activities outside normal work and |
50 |
> family commitments manage a man month's worth of effort over 12 months of |
51 |
> elapsed time. So you're really looking at 360 years to get it done by |
52 |
> volunteers. |
53 |
|
54 |
Frankly I don't think that selling this idea to Novell, IBM or RedHat would |
55 |
be impossible. The amount of value that a common configuration standard |
56 |
would give to Linux is huge. It would make it considerably easier to create |
57 |
configuration tools and thus could drop the TCO of Linux systems quite a |
58 |
bit. And TCO is what Linux in the corporate world is really all about. |
59 |
|
60 |
-- |
61 |
Jani-Matti Hätinen |
62 |
"Oh I'm not drunk. I'm mentally ill" |