jim@haring.UUCP (04/01/84)
A (to some) considerable part of the net is already functioning with compaction, everything to the east of and including decvax and philabs. The trans-Atlantic links to mcvax and most feeds from there to other countries in Continental Europe use a scheme suggested by Armando Stettner and implemented by me, which uses Berkeley compact/uncompact married to the programs to be found in a subdirectory of the news 2.10 distribution. This has been working well since October '83. Average data compaction is around 33%, plus there is lower overhead with UUCP protocol, giving larger savings than at first sight. This scheme is completely separate from the 'normal' batching scheme, and the two can be maintained side by side. It has an option to specify the maximum size of (uncompacted) batch file to create, so it may make several smaller files rather than one large one. This is needed as it the success rate in trying to transfer 50K+ files in one attempt across the Big Puddle is not high. James A. Woods, ames-lm!jaw, makes some comparisons between the Berkeley compact/uncompact and the older pack/unpack. He may be right, the Berkeley ones do seem to take a long time. Their advantage is that they work on standard input/output, which lends itself nicely to this scheme and keeps the complexity of the system down (I shudder at the thought of making the news system MORE complex). There may be other, better ways of doing this, I wouldn't be surprised. But parts of the net are already using a compaction scheme, including two major sites in the USA, and that should be taken into consideration when designing/proposing a new one. Jim McKie Centrum voor Wiskunde en Informatica, Amsterdam ....mcvax!jim