1 |
Ciaran McCreesh wrote: |
2 |
> On Wed, 2 Nov 2005 19:33:37 +0100 Jan Kundrát <jkt@g.o> wrote: |
3 |
> | "Once this tool is implemented and well tested it can be integrated |
4 |
> | into portage." |
5 |
> |
6 |
> can ! will. It might, but don't count on it. |
7 |
> |
8 |
> | GLSA already contains stuff for marking items as valid only for given |
9 |
> | systems, for "injecting" them etc. Why don't use existing code? Why |
10 |
> | duplication? |
11 |
> |
12 |
> Because it's quicker to invent a wheel which is actually round. |
13 |
|
14 |
Reinventing rounder wheels seems to be a common hobby. |
15 |
|
16 |
> | > You think XML magically makes things compatible? Then I suggest you |
17 |
> | > write a GuieXML to Docbook conversion tool, and see how many |
18 |
> | > thousand lines of XSLT it takes. All XML does is move the |
19 |
> | > conversion and parsing problems to a different, more complex level. |
20 |
> | |
21 |
> | I'm not familiar with DocBook, but I doubt I'll need thousands of |
22 |
> | lines of code. |
23 |
> |
24 |
> Oh? Our GuideXML to HTML conversion is thousands of lines of code... |
25 |
|
26 |
Plain wrong, but you have always made it clear that you are not only biased |
27 |
against XML for anything, but also very much XML challenged. |
28 |
Don't worry, some are even worse than you are, worse enough to claim that XML |
29 |
is hard to parse because "XML files from a programming perspective require |
30 |
extra logic to parse. Compare the following key value pair and xml tag pseudo |
31 |
parsing logic for configuration: |
32 |
<tag1>entry</tag1> |
33 |
Hit a >, tag1 as realized tag name, read until <, read ahead one to ensure a |
34 |
closing slash, read until > to get the tag name, compare tag name with |
35 |
previous tag name to see what tag it's closing. store value attached to tag1." |
36 |
|
37 |
Just a short sample against metadata.xml using ruby/dom instead of python/sax: |
38 |
http://dev.gentoo.org/~neysx/metax.rb |
39 |
|
40 |
Discarding XML for the reasons some are using is like recommending key=value |
41 |
flat .ini files because windows used it in the 80s. |
42 |
They have to be parsed as well, and, as opposed to XML, you have to check for |
43 |
unknown keys, double keys, missing ones, then you need some grouping and you |
44 |
introduce [sections], which you have to check as well, no doubles, no missing |
45 |
ones, no illegal ones, then you need a deeper hierarchy and you use |
46 |
key=/path/to/another/file.ini... |
47 |
|
48 |
Both have reasons to be used, neither is a one-fits-all answer. |
49 |
|
50 |
Anyway, this is getting off-topic, and, FWIW, I believe the suggested format |
51 |
is adequate because it is light, easy to write, read, parse, and even |
52 |
transform into XML should one process ever need it. Besides, it is very much |
53 |
standard, if it's good enough for billions of mail and http messages a day, |
54 |
it's probably good enough for us. |
55 |
|
56 |
-- |
57 |
/ Xavier Neys |
58 |
\_ Gentoo Documentation Project |
59 |
/ French & Internationalisation Lead |
60 |
\ http://www.gentoo.org/doc/en |
61 |
/\ |
62 |
|
63 |
|
64 |
-- |
65 |
gentoo-dev@g.o mailing list |