From: greg andruk (gja@meowing.net)
Date: Wed Oct 09 2002 - 21:55:08 CDT
Charles Lindsey wrote:
> I think every "prettyname" needs to have inbuilt into it the identity of
> the authority who created it. In most cases, that could be just the name
> of the hierarchy.
It already has that. I see that Andrew also missed this, so I guess
that I should try to explain a bit more fully.
Earlier in this discussion, there have been a few mentions of groups
organized into logical or administrative collections. This is actually
just a generalization of what we have today; we currently call them top
level hierarchies. Later (as in not now), we can allow collections to
be a bit more arbitrary, but if we define a group's _default_ collection
as being the same the top level of its traditional group
specification, we have something we can work with today without hacking
into Usenet's guts.
It would really be nice if we could avoid going into great detail about
how precisely to implement more detailed collections at this point. I'm
hoping that it will be enough to explain the general idea and leave it
as something to be tackled in a later effort, and that for now we can
just leave a placeholder for such a thing to be implemented. Doing an
advanced version of collections really well would ultimately require
smarter server software and to tackle that pesky authentication stuff,
and we already know that this particular working group is never going to
agree on how to implement auth. So for now, let's work with what's
already there, okay? Yes, I'm begging. We have a terrible tendency to
go off on tangents in this WG, and never manage to get much finished
that way.
>>And yes, it's okay for groups in different administrative domains to
>>give their groups overlapping display names, just like it's okay for a
>>mail server to have multiple users named John Smith. Once again, with
>>feeling: names are not addresses, and these are solved problems.
> Yes, but even a mail server needs some mechanism to disambiguate "John
> Smith".
Right, and since (by default) an administrative domain is essentially a
top level hierarchy, another way of saying that is "display names need
to be unique _within a given top-level hierarchy_." We actually have a
form of this today; alt.lang.intercal and comp.lang.intercal are the
same name under different top level hierarchies.
Under the extended naming scheme, the display name is already tied to
the group specification, just like the m/y posting flag. Adding the
hierarchy name to the display name in wire protocols adds no value, it
is a user agent interface issue. Putting the hieracrchy into the
display name actually would interfere with future collection
possiblities, making it confusing for a group to belong to more than one
collection. [One example of a potentially useful future collection that
doesn't fit well into the current hierarchical structure might be one
for large binaries. But again, that's not something to think about in
minute detail today, it's enough to be mindful of the possibility.]
One newsreader might list group names in a tree format, much as GUI
newsreaders already tend to do today:
Alternative Net
C programming language
Kibology
Fans of Karl Malden's nose
Computers
C programming language
ASUS PC mainboards
UNIX System V bugs
Another agent might put this information into columns:
Name Collection
---------------------------- ----------
ASUS PC mainboards alt
C programming language comp
C programming language alt
Kibology alt
Another agent with a more keyboard-oriented approach might choose to
handle them in an Outlookish sort of way, with some kind of
autocompletion thingy in mind:
C programming language [Alternative Net]
C programming language [Computers]
...and there are obviously many other user interface options. I'm just
trying to illustrate, not provide an exhaustive list.
The hardcore elite minimalist geek can keep on using the traditional
group specifications, and normal people will never notice.