Re: [idn] IDN priorities and requirements

[James Seng <jseng@pobox.org.sg>]

Larry Masinter wrote:
> I think the list of "places where domain names appear" is useful,
> but I wanted to explore your list a litte more deeply.

Great idea.

If anyone is doing such a list, I would like to expand the list to cover
"protocols where domain names are used".
 
> > logs
> 
> I'm not sure what kind of logs have domain names in them, that
> aren't also full of other ASCII material and that are also
> commonly visible to users. Could you give some examples of
> the kinds of logs & the software that writes them?

Quite a few, but most common syslog and apache logs both write FQDN.
you can think of a few other easily...(unless u dont consider
sysadm as a user :-)
 
> > access lists
> 
> You mean like a 'rhosts' file? I suppose that would be one
> kind of application. What other kinds of access lists contain
> host names?

sshd.conf? access.conf? httpd.conf? hmm...
 
> > traces,
> 
> These are different from logs?

e.g. traceroute. other tools like SNMP?
 
> > news papers, magazines,
> 
> I've seen email addresses and URLs printed in newspapers and
> magazines, or else host names that are intended as URLs with
> 'http' implicit. But for the most part, publications use domain
> names as URLs.

agreed. and email too.
 
> > server addresses, ftp addresses,
> 
> These are different from URLs, but often expressed as them.

agreed. hardly anything cant be expressed as url nowadays...

except the other 1000+ protocols inside /etc/services which i dunno
what and how to use them...

do we have?
mysql://foo.bar/ or
sa-msg-port://foo.bar/ or more useful
whois://whois.internic.net/ietf.org

> > zone files.
> 
> I think it's stretching things to say that "zone files" are an
> application where users commonly see domain names.

while it is not an application, it is definately a place where 
labels and domain names appears...
 
> > It does not matter if the user is a simple non-computer expert,
> > hacker, administrator or programmer.
> 
> If any progress is to be made in this area at all, some tradeoffs
> will have to be made. If it were possible to satisfy some of those
> kinds of users but impossible to satisfy all of them, wouldn't you
> want to prioritize?

it depends on what kind of tradeoff in security you are talking about.

if it affects a couple of OSes and some minor applications where we can
inform vendors and fix it, that is 'okay'. if it affects every machines
on the Internet now, we need to hold back just a little...
 
> > They all want to use
> > domain names in a natural way: they want to use the characters
> > of their native language. They do not want to see any encoded into
> > ASCII forms!
> 
> I don't doubt that there is such a desire, but each category of
> software that you want to be compatible with imposes a different
> set of constraints. To make progress, it might be helpful to analyze
> those constraints realistically.

agreed.
 
> While the analysis for I18N of URLs and email addresses has been
> studied, I'm less certain about what kinds of constraints would
> be imposed by the requirement, for example, of allowing I18N names
> in log files. Aren't log files usually stored in the local operating
> system charset?

we seem to have an impossible job :-)

*cheers*

james