Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP Path: utzoo!mnetor!uunet!seismo!husc6!hao!ames!ptsfa!ihnp4!ihlpl!jhh From: jhh@ihlpl.ATT.COM (Haller) Newsgroups: comp.unix.wizards Subject: Re: Size of SysV "block" (really: byte != 8 bits) Message-ID: <2425@ihlpl.ATT.COM> Date: Thu, 23-Jul-87 16:42:12 EDT Article-I.D.: ihlpl.2425 Posted: Thu Jul 23 16:42:12 1987 Date-Received: Sat, 25-Jul-87 11:11:53 EDT References: <857@bsu-cs.UUCP> <326@hubcap.UUCP> Organization: AT&T Bell Laboratories - Naperville, Illinois Lines: 43 Summary: Non-eight bit bytes should die In article <326@hubcap.UUCP>, beede@hubcap.UUCP (Mike Beede) writes: > Seriously--different machines serve different purposes, and so are designed > differently. That is why it is foolish to freeze some design parameter > arbitrarily. I don't see that there is, for instance, a clear argument > against 36 bit words and 9 bit bytes as opposed to 32 bit words and 8 bit > bytes, especially if your application works well with 9 bit quantities. The clear argument against 36 bit words and 9 bit bytes is data communications. Like it or not, data communications have virtually standardized the 8 bit byte. Just try to generate TCP/IP headers on a 9 bit machine, packing all of the data contiguously. Oh, you want to use a communications processor to do that? How many bits is its byte? Look at some of the higher level ISO protocols, and you will find that the basic unit of data is, surprise, surprise, an octet. Oh, sure, there is support for arbitrary bit strings, but even they are padded to octet boundaries. Back in the 60's, and possibly up to the mid 70's, when 7 track mag tape wasn't considered hopelessly obsolete, there was room for argument on what size provided the best 'byte'. However, the de facto standard is an 8 bit byte, which is becoming more and more institutionalized as time progresses. Given that a byte is an important measure, byte addressability becomes important in hardware architectures. Given that our machines operate with binary logic, word sizes are going to be powers of two bytes long, just so that byte addresses can be easily converted into word addresses, which is typically related by a power of two to the memory and bus architecture. Look at the Harris/6 if you want to see what kind of contortions were necessary to provide byte addressability with a 24 bit word size. In summary, I agree that while there was no good technical reason to have an eight bit byte originally, anyone designing a new computer that does not have an eight bit byte will be doomed to market failure. If Univac's 1100 series had taken off better than IBM's machines, I would probably be saying that six bit bytes are the wave of the future. That is not the case. John Haller