Re: Registration of new IBM Char Sets

[tamer@ca.ibm.com]

I agree with Keld; for example EBCDIC char sets are not directly used on
the Internet, however they are used on the servers that run many of the
internet applications. Conversion has to take place between these EBCDIC
char sets and char sets that are used on the internet. Developers need to
know what char sets are used on the server side to plan the right
conversion. Another case is remote printing over the internet; many of the
IBM printers use the EBCDIC char sets in AFP printing (Advanced Function
Printing), and hence internet applications need to know the char set used
on the remote printer.
Additionally, people have used the IANA registry as a reference for char
set names etc. regardless of whether their applications are for the
internet. I think the scope and usefulness of the registry is wider than
IETF and the internet.
Best Regards,
______________________________________________________
Tamer Mahdi
Globalization Center of Competency
IBM Toronto Lab
81/979/895/TOR
Phone: (416) 448-2680, Fax: (416) 448-2747
e-mail: tamer@ca.ibm.com




Erik van der Poel <erik@netscape.com> on 05/30/2000 05:51:43 PM

Please respond to Erik van der Poel <erik@netscape.com>

To:   Keld Jørn Simonsen <keld@dkuug.dk>
cc:   Tamer Mahdi/Toronto/IBM@IBMCA, ietf-charsets@iana.org, iana@iana.org,
      iana@isi.edu, Antoine Leca <Antoine.Leca@renault.fr>, Harald Tveit
      Alvestrand <Harald@Alvestrand.no>
Subject:  Re: Registration of new IBM Char Sets




Keld Jørn Simonsen wrote:
>
> The idea was to document it for applications to be used at a platform
> but not on the wire. And this was deemed needed as IETF specifications
> are also about the things that happen with the protocols at
> the platforms.
>
> On the other hand, it was decided that only some selected charsets
> should be used for interchange, on the wire. Maybe we should add
> that as a property to each of the registrations, that is,
> "recommended for use on the wire". I believe some of these
> recommendations are already done in the rfcs specifying the
> protocols.

The charset names are quite useful. In fact, Mozilla uses them to
identify converters that convert to and from Unicode. Mozilla uses
Unicode internally, and converts to and from whatever charsets are used
in the underlying software (OS, etc) and on the Net.

But as things stand, there is not very much guidance in the use of the
many charset names in the registry, and people may feel free to use any
of them on the wire, when the working group's intention was to minimize
the number of charsets on the wire.

So I would tend to agree with you that we should do something about
this, but I don't know whether an additional property in the registry is
the best way to go. Perhaps some protocols would prefer to be very
strict and only allow a very small number of charsets, while other
protocols would prefer to be quite liberal, allowing a larger variety of
charsets.

Comments?

Erik