i18n on Win32

["William A. Rowe, Jr." <wr...@rowe-clan.net>]

Here's the analysis,

  We can port and drop in iconv-1.1, iconv-extra-1.1, and iconv-rfc1345-1.1
by Konstantin Chuguev from FreeBSD.  License of sources is essentially
compatible, but includes the advertising clause, requiring that we cite him
and include the disclaimer (which is essentially our own disclaimer.)

  The other/better alternative would be IBM's ICU package, but it isn't going
to be a direct and simple thing, since I find no charset <-> charset funcs,
only conversion into and out of Unicode.  The IBM Public License, of course, 
is more compatible and doesn't contain the advertising clause.  I would also 
put more faith into IBM and the ICU colaborators' efforts to have their tables 
right, and correct and expand them very quickly.

  Jeff, had you considered ICU when you started your i18n work?  If so, what
were the shortcomings (e.g. the missing charset <-> charset functions) or
did you find these functions?

  Implementing xlate within ICU is probably no great challenge, and ICU is
already ported to Win32, so does anyone care, or should I stay with iconv?

Bill

[Either way, this will wait for the next alpha/beta to be deployed.]

Re: i18n on Win32

[Jeff Trawick <tr...@bellsouth.net>]

"William A. Rowe, Jr." <wr...@rowe-clan.net> writes:

> Here's the analysis,
> 
>   We can port and drop in iconv-1.1, iconv-extra-1.1, and iconv-rfc1345-1.1
> by Konstantin Chuguev from FreeBSD.  License of sources is essentially
> compatible, but includes the advertising clause, requiring that we cite him
> and include the disclaimer (which is essentially our own
> disclaimer.)

(naive) perhaps the advertising clause is a legacy from the old BSD
license and he would be willing to "upgrade"?

>   The other/better alternative would be IBM's ICU package, but it isn't going
> to be a direct and simple thing, since I find no charset <-> charset funcs,
> only conversion into and out of Unicode.  The IBM Public License, of course, 
> is more compatible and doesn't contain the advertising clause.  I would also 
> put more faith into IBM and the ICU colaborators' efforts to have their tables 
> right, and correct and expand them very quickly.


> 
>   Jeff, had you considered ICU when you started your i18n work?  If so, what
> were the shortcomings (e.g. the missing charset <-> charset functions) or
> did you find these functions?

I never considered ICU; I had never heard of it.  Code I'm familiar
with has all used iconv().  Importantly for Apache development,
iconv() was available and well-supported on enough platforms for me
already (OS/390, Solaris, recent glibc on Linux).  It is nice to use
iconv() on platforms which support it natively because administrators
can add support for their own special translations once and everything
works with it. 

>   Implementing xlate within ICU is probably no great challenge, and ICU is
> already ported to Win32, so does anyone care, or should I stay with
> iconv?

There is nothing particularly good about iconv() other than the fact
that on some platforms most everything uses it, making it is easy for
administrators to add support for special translations since adding to
the iconv() database instantly upgrades most apps.

On platforms where neither ICU nor iconv() is native, it probably
doesn't matter to the end user (unless one of them doesn't work well
or doesn't have an easy way for users to add tables or custom
translation logic).

-- 
Jeff Trawick | trawick@ibm.net | PGP public key at web site:
     http://www.geocities.com/SiliconValley/Park/9289/
          Born in Roswell... married an alien...

Re: i18n on Win32

[rb...@covalent.net]

>   We can port and drop in iconv-1.1, iconv-extra-1.1, and iconv-rfc1345-1.1
> by Konstantin Chuguev from FreeBSD.  License of sources is essentially
> compatible, but includes the advertising clause, requiring that we cite him
> and include the disclaimer (which is essentially our own disclaimer.)

We have other code in Apache that includes the licensing clause.  Although
we haven't added any since we removed our own licensing clause.  :-(

>   The other/better alternative would be IBM's ICU package, but it isn't going
> to be a direct and simple thing, since I find no charset <-> charset funcs,
> only conversion into and out of Unicode.  The IBM Public License, of course, 
> is more compatible and doesn't contain the advertising clause.  I would also 
> put more faith into IBM and the ICU colaborators' efforts to have their tables 
> right, and correct and expand them very quickly.

I don't know if I agree with the correctness argument.  Remember FreeBSD
is VERY popular in other countries, so I suspect that those tables are
always up-to-date.  It looks like a real problem that you can't do
charste <-> charset translation with ICU.  But, I don't have much
experience or interest in i18n right now, so I may be making too big a
deal out of this.

>   Implementing xlate within ICU is probably no great challenge, and ICU is
> already ported to Win32, so does anyone care, or should I stay with iconv?

My own opinion is that you know what you are doing, please do what is best
for the code.  :-)  Nobody else has done your research on these options,
so we are likely not the best people to ask this question of.

Ryan

_______________________________________________________________________________
Ryan Bloom                        	rbb@apache.org
406 29th St.
San Francisco, CA 94131
-------------------------------------------------------------------------------