Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP Path: utzoo!watmath!clyde!cbosgd!ulysses!bellcore!decvax!decwrl!pyramid!pesnta!hplabs!hao!nbires!isis!aburt From: aburt@isis.UUCP Newsgroups: net.mail,net.news Subject: Re: Data compression to lower phone Message-ID: <762@isis.UUCP> Date: Thu, 5-Jun-86 00:39:11 EDT Article-I.D.: isis.762 Posted: Thu Jun 5 00:39:11 1986 Date-Received: Sat, 7-Jun-86 07:02:08 EDT References: <327@spdcc.UUCP> <8200002@nucsrl> <2369@phri.UUCP> Reply-To: aburt@isis.UUCP (Andrew Burt) Organization: University of Denver Math and Computer Science Lines: 50 Xref: watmath net.mail:1658 net.news:4954 In article <2369@phri.UUCP> roy@phri.UUCP (Roy Smith) writes: >In article <8200002@nucsrl> gore@nucsrl.UUCP writes: >> [...] a site that passes news articles to other sites could just pass >> on the compressed form (and tack on a compressed form of articles >> originate from it). > > Funny, I was just talking with someone this afternoon about this. >There are a few problems. The toughest one is that every article that >passes through here has to get "phri!" prepended to its path line. No way >to do that without going through the uncompress/compress cycle (at least, >not that I can see). Pity. I see one, but it ain't pretty. Modify the next version of news software so that rather than recompressing the news to send it instead sends along the sitename to prepend on the other end when unbatching. (Controlled by a 'sys' file flag, of course so that you don't send wrong paths to those who can't handle it.) E.g., site A creates a compressed batch of news and sends to B. B sends along the compressed file to C, along with a note saying to prepend B! to the path when C unbatches. If C decides to send the compressed batch on to D, it sends a note saying to prepend C!B!, etc. (-: And hey -- this might a few cents on the phone bills too: In our neck of the woods, Path lines seem to average 86 characters (minus the "Path: " and \n). A random sampling showed about 60 articles per 50K batch file. In the best case, the batch we get would be the same batch that left the first backbone site the article encountered (which probably would run csendbatch to repackage the little batches it got into 50K batches to send onward). So figure that half of the path, or 43 characters, is sent once rather than 60 times for that batch. If compress reduces the batch to 40% it's size, we have (43*60*.4 - 43) or 989 bytes saved per 50K of compressed news, about 2%. Now, this isn't the problem the idea was intended to solve, but it might have small positive financial side effects as well. The idea definitely breaks down when site C doesn't want net.foobar from B but gets it transmitted anyway. However, as an option where the conditions are right, it might work ok. :-) -- Andrew Burt isis!aburt or aburt@udenver.csnet