🚀 go-pugleaf

In article <1il974INNq3a@rodan.UU.NET> revell@uunet.uu.net (James R Revell Jr) writes:
>} I heard that such mechanism was available in a commercial
>} product (Dynafeed for ClariNet?).  Isn't it time to give the
>} concept wider exposure?  Does anyone know how it is done and how
>} [un]practical it is?

Dynafeed never was a commercial product is the sense of charged for.
It is free.  If a professional newsfeed like uunet wants to use it, we
ask that they help us feed ClariNet news in exchange.  That's it.

>
>Note that dyamically updating the sys file is quite different than
>everything Dynafeed does.  Dynafeed does allow you to control what
>groups you receive from the site fed and to some degree even the
>articles within the group, but Dynafeed uses a completely different
>method of feeding which I do not believe is efficient enough to be used
>on a large scale.  Sites running a number of such feeds are encouraged
>to comment.

No, according to what I know of INN, it can read a sys file of arbitrary
complexity -- ie. a plain list of newsgroups -- and internally it generates
a plain list of newsgroups.  This is what dynafeed outputs and sends upstream.

That list of newsgroups can of course be used with the dynafeed feed program,
but it is a fairly simple change to turn it into a sys line that INN can
use efficiently, and in fact when I get around to using INN, I will put
an option on the dynafeed arbit program to output an INN style sys entry.

As to whether the "feed" program is efficient, I think it is reasonably
efficient, and If I bothered to move over Xref code from newsclip
(which is also now free, though it started life as evalu-ware) it would
be as efficient as any other scheme in terms of disk reads.

Right now the only inefficiency in the "feed" program is that it reads
crossposted articles multiple times and junks them all but one of the
times.  This is a wasted file open and block read.   If it bothered to
remember what it had seen in memory through xref lines, it would not
have this problem.  It is a very simple program, though, and I wanted to
keep it that way.

Of course, this efficiency is only on uucp and other batched feeds.  It
is indeed less efficient on an nntp feed, particularly in contrast to
inn.

I think it would be more efficient than C news on a batch uucp feed if
I put in support for Xref lines.

In C news we have a file open, write and close for each site when an
incoming article batch arrives.   Then, when it is batching time, we
open and read all the outgoing articles and write them to outgoing
batches.

Dynafeed's "feed" has no overhead when an incoming batch arrives, and
when it is time to batch it opens and reads the outgoing articles and
writes them to the batches, with the wastage of duplicate opens due to
the simple xref scheme.   Without that wastage, it does less disk I/O
than C news.

Dynafeed of course works in a software tools approach, with several
programs in the toolset.  The readership measurement tools, which produce
a list of what's being read at your site and downstream, then send that
list upstream, could easily be used with another feeding program upstream.

Likewise the feed program can be used for static feeds, even if requests
are not coming from downstream.
-- 
Brad Templeton, ClariNet Communications Corp. -- Sunnyvale, CA 408/296-0366