From: Brad Templeton (brad@templetons.com)
Date: Fri Mar 03 2000 - 13:03:47 CST
On Fri, Mar 03, 2000 at 12:39:58PM +0000, Charles Lindsey wrote:
> In <20000302174159.C6930@main.templetons.com> Brad Templeton <brad@templetons.com> writes:
>
> >I was beginning to believe that by defining the xref requirement you even
> >wanted to forbid replace-in-place. I'm fine with the system having the
> >option of either.
>
> Well you have to be careful that all newsreaders will work with all
> servers (or all NNTP servers in practice). I have not thought through
> whether your method runs into any problem there.
>
> It is certainly the case, with the system as currently envisaged, that
> users will not see the benefit of the new feature until BOTH their local
> server and their own newsreader have implemented it. Servers that
> implement your form may well enable old newsreaders to see the new
> functionality. But yours is a bigger upheaval to the server
> implementation.
Actually, I thought it was simpler! The only upheval would be the
midly complex suggestion I had for how to do in-place replace for long-term
streams. If you do only short-term replacements, it's pretty simple,
other than the added code to rewrite the overview entry for a given
article number. Not free, but not hard.
I think we would be better off if we had the servers do a bit more work
on the long-term chains to save the newsreaders doing *any* work, but I'm
not going to beat my head against the wall here about it.
As for the value of having the users of old newsreaders gain benefit,
I can't stress that enough. It will take years, many years before
everybody has a new newsreader, and unless we deliberately break things,
we will never have everybody using one. Servers will upgrade much, much
more quickly. The time before "90% of users are reading from a recently
coded server" is vastly sooner than the time "90% of users are using a
recently coded newsreader" and both of those are way ahead of 90% of users
using both.