Path: utzoo!censor!geac!torsqnt!news-server.csri.toronto.edu!cs.utexas.edu!usc!julius.cs.uiuc.edu!apple!bionet!hayes.ims.alaska.edu!accuvax.nwu.edu!nucsrl!telecom-request From: ms6b+@andrew.cmu.edu (Marvin Sirbu) Newsgroups: comp.dcom.telecom Subject: Re: Prodigy Responds to E-Mail Criticism Message-ID: <15167@accuvax.nwu.edu> Date: 2 Dec 90 19:34:19 GMT Sender: news@accuvax.nwu.edu Organization: TELECOM Digest Lines: 31 Approved: Telecom@eecs.nwu.edu X-Submissions-To: telecom@eecs.nwu.edu X-Administrivia-To: telecom-request@eecs.nwu.edu X-Telecom-Digest: Volume 10, Issue 861, Message 7 of 12 The Prodigy system is designed like a multi-level memory hierarchy. Information is stored initially at some nationwide location. As it is demanded in a particular city, it is copied to the city node and cached there. Thus subsequent reads do not require a transfer from the national headquarters to the regional node. (Cacheing is also done in the PC, but that is irrelevant to the point of this message) A bboard post, if widely read, will be copied from the national host to each regional, and then read from the regional many times. Thus, every transfer from the national to the regional is "amortized" over multiple reads. I infer from the information supplied by Prodigy that all individually addressed mail goes up to the national host and then down to the regional for delivery to the recipient. I also suspect for efficiency, the regionals are designed only to do object cacheing, independently of the type of object. If so, it would be a fairly radical change to reimplement mail so that mail objects with multiple destinations are not replicated at the national host, but in a two step process that would send one replica to each region where there are addressees, and the region would then replicate the object for each addressee. This would require the regionals to do more than object cacheing: they would have to examine the content of the object. If replication does happen at the national level, then, indeed, a multi-addressed message is much less efficient than a bboard post. Marvin Sirbu Carnegie Mellon