From: greg andruk (gja@meowing.net)
Date: Sun Oct 06 2002 - 10:37:33 CDT
Erland Sommarskog wrote:
> However, the use of UTF-8 only requires user agents to be updated
False. Servers would require modifications to their moderated group
handling and control mesage processing
> (and
> contrary to what you say, not in all cases). Solutions with prettynames
> or use of other charsets also require changes in server software.
Tehrre is a very big difference between the server changes required to
support an encoded Newsgroups: header and to support a name extension.
The in-band approach requires server changes to regain broken
interoperability. The out-of-band approach retains intreroperability
without the need for flag days to support the first moderated group with
a broken name.
> The one thing you achieve with solutions where the on-the-wire name
> of the newsgroup is pure ASCII, is that you can say that it will not
> break any technical interoperability without even looking.
Hardly; people have been looking at it for 22 years. ASCII newsgruop
specifications are very well understood at this point.
> The suggestion
> to use UTF-8 requires the effort to investigate, but the result of
> the investigation is positive to a great extent.
Except when it's negative, and it is already known to break moderated
newsgroups and gateways. Unescaped UTF-8 does not interoperate with the
existing news infrastructure, which, like it or not, does include the
mail infrastructure.
> But, alas, without updated software, the extensions are not very useful
> at all.
This is true of all changes to the way newsgroups are named. ASCII
newsgrouip specifications, for all their limitations, do at least
provide visible strings with essentially universal support.
It is already know that, for example, Windows software will
automatically and invisibly change other chanracter sets during
cut-and-paste operations. This cannot be fixed without updating software.
> And if they are not useful, people will not use them. And if they
> are not used, why waste the effort of implementing them?
The leading newsreaders jumped on board immediately with test
implementations. It's simply not worthwhile for everyone to jump into
these things unless and until a real standard emerges, but client
writers will add whatever becomes the standard.
> True, when we discuss a the new group in the se.* hierarchy (not that we
> do this anymore), we would have to determine both a regular name and a
> "prettyname". But the regular name would be at least as important as
> the prettyname,
For the near term, yes, it would remain important. Disconnecting the
name space from the address space does, however, allow for the current
hieracrhical structure to become less and less important over time.
Merely extending the character set leaves that problem unsolved in
perpetuity.
> and all the concerns we have today how a name which
> includes едц will be interpreted when replaced with aao would still
> remain, and we would in some cases still have to bend over backwards
> to find a name which does not have any such problem.
It becomes much less contentious when the newsgroup specification is
reduced to an address specification, and is freed of its ill-fitting
alternate definition as a newsgroup's name.
[snip mail examples]
> And that is a feature that could be implemented locally in your newsreader,
> without any need for support elsewhere, nor in servers nor in RFCs.
Which means, of course, that all the proposed incompatible changes made
within the Usefor WG are unnecessary. That is pretty much the point,
that the article header is the wrong place to make changes.