From: Kai Henningsen (kaih@khms.westfalen.de)
Date: Wed Nov 17 1999 - 01:01:00 CST
chl@clw.cs.man.ac.uk (Charles Lindsey) wrote on 16.11.99 in <FLACpJ.9yD@clw.cs.man.ac.uk>:
> In <7SnpO2NHw-B@khms.westfalen.de> kaih@khms.westfalen.de (Kai Henningsen)
> writes:
>
> >This is a problem for restricted feeds. If (this is some years back) I
> >want a feed of exactly comp.os.linux, I wouldn't see the newgroup messages
> >advertizing the split, though it's pretty obvious that I'd want to see
> >them.
>
> According to our present draft, if you do
>
> Control: mvgroup comp.os.linux comp.os.redhat
>
> then the Newsgroups header MUST include at least comp.os.linux and
> comp.os.redhat (it MAY include more). Is that not sufficient for your
> purpose?
I was talking about the generic control propagation problem, not
specifically about mvgroup. Given that mvgroup is new, I have no
operational experience with it :-)
I certainly want to see all newgroups in comp.*, even if I do not have a
full comp.* feed.
I think we have a significant change from the time when control message
propagation was originally designed.
Back then, most feeds were of the "comp.*,!alt.*" form, that is, one
decided on the general hierarchies one wanted (or so I gather). It was
reasonable to just restrict control message propagation the same way.
These days, there are also a lot of very restricted feeds that only take
specific groups. It is no longer reasonable to have the same restriction;
people taking some groups from comp.* may very well want to see all
control messages for comp.* so they can keep an accurate group list of it,
so they can decide how to change their feed.
That is, some feeds today work much like newsreaders do: there is a
complete group list and a much smaller list of actually subscribed-to
groups.
This is not just theory; news.westfalen.de and all *.westfalen.de sites
getting a feed from it work that way. Curently, we do some of it by having
"control*,!control.cancel" patterns. That's suboptimal, however.
MfG Kai