clewis@ferret.ocunix.on.ca (Chris Lewis) (06/15/91)
In article <1991Jun12.161500@mccall.com> tp@mccall.com (Terry Poot) writes: |In article <2121@ecicrl.ocunix.on.ca>, clewis@ferret.ocunix.on.ca (Chris |Lewis) writes: |>A couple of additional things I've thought of: |.... |> 2) There's no need to enable this on all sites. There could always |> be a question in the build procedure asking whether it need |> be enabled or not. If you are a leaf node, then there's |>relatively |> little utility in turning it on. If you're a site with a big |> fanout or a regional chokepoint, you should have it on. |Unless you are strictly a leaf, it should be enabled. If you drop a message |and don't tell anybody, nobody may ever know. If you are a feed for even |one site and he gets some bogus software, he should be told. This should be |low overhead, since the error report is only ever generated when a message |is dropped. You're probably right. There are ways of being a little less rabid in enabling the stuff, but difficult to implement. |> 3) One thing that could greatly reduce the loading, is have the |> news software only send the error article *back* to the site that |> fed it the erroneous one. Not any other sites. This has the added |> advantage that regions fed thru a small number of gateways (especially |> only you, or if all the gateways were similar C news sites) would often |> *not* propagate the article to the rest of the net - it would only |> be on the machines in the area "beyond" the gateways. |Unless you are a choke point for a subnet, this won't work, as the message |will flood around you. Who cares? In the worst case, it cuts the transmission loading down to precisely one emission per site as opposed to one per outfeed... Which would make Kent a little happier (in addition to, of course, not doing this at all for stale news) |In the cases it does work, it keeps the error |reports from getting to the central email sites, reducing the likelihood |that a responsible party will see the notification. Who cares about "central email" sites? What you really want to do is make sure that the originating site or some subset of the originating site's neighbors sees it. Which would be extremely unlikely except in strictly unidirectional feeds (though, this could be bypassed by special sys entries) |I personally would prefer a per-system limit with a larger interval, say |one error report per system name per day. Much better - your suggestion on how to do this efficiently removes my objections. -- Chris Lewis, Phone: (613) 832-0541, Domain: clewis@ferret.ocunix.on.ca UUCP: ...!cunews!latour!ecicrl!clewis; Ferret Mailing List: ferret-request@eci386; Psroff (not Adobe Transcript) enquiries: psroff-request@eci386 or Canada 416-832-0541. Psroff 3.0 in c.s.u soon!