Re: [xml] Building libxml2 Windows binaries against win_iconv instead of GNU libiconv



Hello Tor,

As far as I am concerned, that is perfectly okay. However, before I can say yes or no, I must consult the mailing list, so I forward this message there.

See, according to what we see on the mailing list, the number of those who use libxml2 on Windows is quite large. Most of them use libxml2 in a more critical missions than I do. My XML files are all in plain 7-bit ASCII, using entity references whereever needed, and I control all of my XML. Most of the others aren't so lucky, they need to process markup of all sorts, very little of which they have control of. Many of them are likely not to have control over which codepages are installed on their Windows systems. Therefore I must let them have a say in this, it affects them much more than it affects me.

To all users:

Here we have a choice. Use GNU libiconv, or win_iconv on Windows. Using GNU libiconv means having a libxml2 which handles all encodings known to mankind at the price of depending on another library, namely libiconv. Using win_iconv means having a libxml2 which handles everything your setup otherwise handles, with the benefit of not being dependent on an additional library.

As Tor rightly pointed out, most Windows setups should be able to handle all that they need and if they don't, they are screwed anyway, so the win_iconv way should work out. However, in my history, I have worked for several large companies and I can say that their setups are far from optimal and there is little one can do about it, unless one is a CEO, or close.

I herewith leave the decision to the user base of libxml2 on Windows. If win_iconv it should be, I'll gladly adjust everything to make it that way. Say your opinions, oh you libxml2 users, show some democratic tendencies :)

Ciao,
Igor



On 25/01/08 15:38, Tor Lillqvist wrote:
(Seems to take awfully long for my subscription to xml gnome org to be
processed, so I send this to just you instead. Feel free to forward to
the list.)

I have a suggestion: I think it would be a good idea to in the future
build the libxml2 Windows binaries against the "win_iconv"
implementation of the iconv API instead of GNU libiconv.

win_iconv is a quite small static library (just one smallish source
file actually) that can be built directly into the libxml2 DLL. It
uses the Win32 API to do the conversions. It removes the dependency on
the rather large GNU libiconv iconv.dl.

It is by Yukihiro Nakadaira, from
http://yukihiro.nakadaira.googlepages.com/#win_iconv . I have a small
repackaging that contains the single source file, a trivial iconv.h,
and a (mingw) library at
http://ftp.gnome.org/pub/gnome/binaries/win32/dependencies/win_iconv-tml-20071205.zip
. (It is trivial to compile also with MSVC, just one source file with
no dependencies.)

As far as I can see, win_iconv works fine as a replacement for GNU
libiconv. It supports only those encodings for which the code pages
are present on the machine at run time, but I think all Windows
machines should have all codepages that are relevant in practise
installed by default. I build my current GLib and gettext distribution
against win_iconv, and they seem to work fine.

--tml





[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]