Path: utzoo!yunexus!unicus!craig From: craig@unicus.UUCP (Craig D. Hubley) Newsgroups: comp.std.misc Subject: Re: Open Software Foundation Message-ID: <2776@unicus.UUCP> Date: 21 Jun 88 21:25:49 GMT Article-I.D.: unicus.2776 References: <5694@potomac.ads.com> <782@dlhpedg.co.uk> <1040@cfa.cfa.harvard.EDU> Reply-To: craig@Unicus.COM (Craig D. Hubley) Organization: Unicus Software Inc., Toronto, Ont. Lines: 92 >> >> >> DEC SEES NEW UNIX VERSION BY AUTUMN >> by Terry Ernest-Jones >> >> "A version of UNIX suuporting the first set of Open Software >> Foundation standards has been promised by DEC for this autumn, >> as AT&T pours scorn on its rival's prospects. >> >> "It is ironic that DEC should be the first foundation member to >> declare support of its standards, since DEC's Ultrix ... was >> rejected by other members as the basis for the operating system. >> Instead a future version of IBM's AIX was chosen. Now, as I understand it, the OSF standard is to be POSIX (IEEE Unix), by god, and IBM is committed to producing a POSIX-compliant AIX that the OSF will license on "reasonable, stable, terms", along with everything else that they plan to produce. I had assumd that nothing would be AIX-dependent, since POSIX was specified and the point of POSIX is to not be UNIX or AIX or ULTRIX. >If this is a quote from an AT&T source it is bizarre because DEC Ultrix >is based on AT&T licensed Unix code (principally BSD 4.x) and therefore >could not be used for an OSF Unix clone product since the whole point is >to write a suitable Unix clone from scratch that has not one line of >AT&T code and therefore requires no AT&T licensing. The IBM product is >a clone, and fills the bill, at least as an OSF starting point. The I thought that AIX was also a BSD port. (Damn it, where did I put those dusty old RT specs ?!?) >only other non-AT&T choice was Apollo Aegis, I believe. Even Apollo >has abandoned Aegis (now using a Sys V Unix I believe) as their >Unix-like product. (Apollo may or may not be calling the product Aegis >but they have moved away from their own clone code to AT&T code and >will presumably move to the OSF clone code when it is ready). I think the more interesting question is what *protocols* are going to be standardized on, not which implementation of POSIX is going to be used. Apollo may yet turn out to be a big winner because their NCS is far superior to the Sun and IBM equivalents. DEC wanted to go ISO in a big way, but was tied up by the fact that the standards simply aren't complete, and thus there is DECNET. NCS actually translates between SNA, NFS, DECNET, and any other protocol that runs on an Ethernet. It has a nifty language called NIDL in which RPC "stubs" are compiled, that look for services at runtime using a location broker, and doesn't get upset that there are several RPC types running on a single machine, or that it has to go to a backup service. Parts of NCS are already PD, and Apollo was pushing it as a standard. For political reasons (Apollo is the smallest of the seven companies, Apollo is effectively a vertical market company, Apollo will not also be the Unix implementor, NCS will work with everything equally well) it seems like a shoo-in. Likewise in the AT&T/Sun world, the silent partner Xerox has already contributed licensing to Open Look and Sys V release 4 will have XNS support. XNS is probably the second-best "universal" protocol presently implemented. It too has location brokers (called Clearinghouses) and a stub definition language (Courier, I believe), but doesn't work so well with all of these other systems. Since this smaller group of three companies can set their own standards, this is less of a problem. Compounding the XNS momentum is that Xerox's Viewpoint software will be, when ported to SPARC, the office automation system of choice, and it runs entirely under (over) XNS. Two-bit prediction: OSF produces POSIX version and supports NCS strongly, if only for political reasons. Although IBM's code will be decried as brain-dead, it will provide the accountability factor that all MIS managers need, and will replace System V as the "standard" Unix. Sun/Xerox coordinate themselves much more closely, and System V release 4 with XNS will become the "gourmet" system, replacing BSD as the "hacker's standard". AT&T will continue to create new "application binary interface" standards weekly until the phrase is as useless as "standard". By trying to keep Unix as their "strategic edge" and at the same time claiming that it is available to anyone, they will shoot themselves in the foot, as they did with the 3B series. Smaller vendors such as Apple, Commodore, and Atari (in that order) will release decent low-end Unix machines (Apple already did, Commodore has one in prototype that they are wrangling with AT&T over, Atari will keep claiming to have one until Unix is obsolete, at which point they will produce it happily: that's what Jack Tramiel did with CP/M for the 64 when he was running Commodore) and won't commit to one path or the other until the software is mature, but I expect them to go POSIX. So, all told, that's not so different from today's situation, except that micros will go Unix and all these silly protocols should be able to talk to each other. Of course, if any of these things doesn't happen, the above predictions are to be considered line noise. -- Craig Hubley craig@gpu.utcs.toronto.edu