mass at akuma.org
Thu Oct 17 17:42:21 CDT 2002
On Tuesday, Oct 15, 2002, at 14:55 America/Denver, Peter Saint-Andre
> On Tue, 15 Oct 2002, David Waite wrote:
>>> Can we just make this a more general assertation then? An xml:lang
>>> attribute SHOULD appear on the stream:stream header, otherwise an
>>> implementation/installation specific default will be used. If an
>>> xml:lang attribute appears on an info-query or message element, its
>>> value supersedes the default of the stream. Does this make sense, or
>>> I missing the point?
>> Thats fine, but it means three things:
>> 1. A compliant server implementation will have to understand the
>> xml:lang attribute on each XML chunk and add them between the parsing
>> the client input and redirection out over an interserver connection,
>> since the default stream:stream language may not be the same.
> Are you saying that a server would need to add the xml:lang attribute
> every chunk? Under what circumstances? The only sense I can make of
> statement is that I may have one default xml:lang on the stream to my
> server whereas my server has a different xml:lang setting on its
> connection to another server. But would a server-to-server stream have
> xml:lang attribute? If so, why? If it did, would that xml:lang
> override in
> all instances? Am I missing something here?
The xml:lang requested by a user, be it per stream or per packet, has
to be preserved across server boundaries. If I request a japanese
locale, this should apply to all entities I communicate with, not just
with the ones on my locale network.
So yes, every chunk which does not match the xml:lang for the
containing document needs to have the xml:lang attribute appended. A
simplistic server implementation might just always add them, while a
more complicated server implementation might try to choose a default
locale for a server to server location that matches the largest
percentage of the traffic over that connection, and remove xml:lang
stamps from chunks which match the default locale.
One way around this really is to say that there is a single,
standardized, default locale which is incapable of being overridden.
This would probably be (US) English based on legacy systems. Then the
lack of an xml:lang attribute always means the exact same thing, and
there is no addition or removal of this attribute based on differing
default locales across delivery hops.
>> 3. The server may stamp its default locale in its stream:stream header
>> to the client. If it does not, it would need to indicate the locale of
>> every XML chunk sent to the client, or the client would have no way of
>> knowing this default locale.
> Then it might be best for the server to indicate its default locale and
> for the client to indicate its default locale, in the respective
> from server to client and client to server.
More information about the xmppwg