Letters to the Editor

Reader Comments and Questions

Send your letters electronically to pubs@uniforum.org. Please include your city and state of residence. We reserve the right to edit letters for space and style considerations.

Unix By Any Name

I write in response to Philip Johnson's article ("The Analyst's Couch," Mar. 1996), which presents his view of the UNIX 95 trademark. In some important areas his facts are incorrect and (perhaps) therefore his conclusions are flawed--or at the very least, capable of a very different interpretation. Mr. Johnson has a responsibility to ensure that his facts are accurate and presented without bias and then argue his point from that position.

UNIX 95 (a space and no apostrophe, please!) is the trademark used on products that are guaranteed by the vendor to be conformant with the Single UNIX Specification. Both UNIX 95 and the Single UNIX Specification are current names for different things: the former relates to a product, the latter to a specification.

Mr. Johnson states that "any product that passes the UNIX 95 certification may call itself UNIX." This is not the case. "Passing the test" on its own buys a vendor nothing. The X/Open brand (in this case, the UNIX trademark) can only be applied to a product when the vendor guarantees in a legal contract that the product conforms (and will continue to do so) to the specification. If a discrepancy is found, it must be fixed, even if a test has been "passed." X/Open does not operate a certification process.

The very high brand awareness of the UNIX trademark cited in the article is the reason for using UNIX 95 for vendors' products. I doubt that even those citizens of Iowa who know about UNIX systems would recognize Posix 95 or XPG5--whatever flavor of alphabet soup that might be! We are using the same analogy as the commodity bleach in the article. In the case of the bleach, the very sameness of all competing products--or even operating systems--brings the basic value. Competition is not about the specification for the product, it is about the ease with which it is available, meets my needs, is affordable and does the job I need done. Won't it be nice when UNIX systems all conform to the same specification and vendors compete on quality of product, service, support, reliability, etc.?

As the person responsible for the UNIX brand, my vision is that MVS, OS/400, et al. do implement the Single UNIX Specification. This can only be to the benefit of buyers, who may then realize the benefits spelled out above. Why the objection to the support of open interfaces in what have, to date, been proprietary products? Provided that they implement the Single UNIX specification correctly and usefully--and the X/Open UNIX brand ensures that they do--then I am inexorably drawn to the conclusion that the animal before me that quacks, waddles to the water and swims away is indeed a duck. Or in this case, if it carries the X/Open UNIX brand and is labeled UNIX 95, it is indeed a UNIX system. And, if five years down the road there is MVS UNIX, Digital UNIX, Reliant UNIX, Solaris UNIX, et al and they are all guaranteed to conform to the same specification, then I have realized the vision and the buyers will be happy.

So finally what will the UNIX trademark mean in the future? It will be the badge for a leading-edge operating environment that is vital and responsive to the needs of the market that is supported by the major vendors worldwide.

Graham Bird
The Open Group
Menlo Park, CA

Philip Johnson is absolutely right! Today, everything seems to be "open," even proprietary systems. Those third-party products already exist. For instance, OpenNT from Softway allows an NT system to become fully Posix conformant (Posix.1 and Posix.2). In fact, it is the same environment required by the U.S. government for all operating system purchases as specified by FIPS 151-2 and FIPS 189. More important than conformance is the strategy behind it. Currently, it is cheaper to buy an NT System with OpenNT than to buy a "real" Unix system.

Novell was right about marketing UnixWare: the right price with the right features; unlimited users, C compiler and two-processor support. It seems that Novell was going back to the old good days of Unix, when everything important was already included in the operating system. But now that is no longer the case. SCO took over UnixWare, and the first thing it did was to increase the price. It will not be surprising if it takes away the C compiler and other neat things, as it has done with its own OS.

The big picture is why buy an overpriced "real" Unix system, when instead I can buy a Unix clone that can run my Unix apps and my Windows apps at the same time? Please do not misunderstand me; I am a Unix "diehard." I love Unix, and I believe it is the best solution to any computing problem, but I do not agree with the strategies of the Unix vendors.

If we want the name Unix to mean something in the future, we have to rethink our strategies about it. When the Unix competitors are so good, how can we justify the high price tag?

Selim Miselem
Miami, FL

Philip Johnson and perhaps UniForum in general have missed the mark once again. I agree, NT and everything else that manages to comply with Spec 1170, Posix or some other open systems criterion is not Unix per se. I remain baffled about why Linux is repeatedly overlooked and excluded from the scope of UniForum commentary; this functionally is Unix. I run Linux with X, Motif, GNU utilities, Xview and all the usual Unix bells and whistles. It functionally is almost identical to most commercial Unix variants I've used. The Caldera Network Desktop and Redhat Commercial Linux are functional, stable products. Caldera is about to offer a port of Wabi and TriTeal a port of CDE, and WordPerfect for Unix and Mathematica run (native) on Linux--the list goes on and on.

Now I ask you, is this Unix?

Steven R. Jones
Virginia Beach, VA

Horsepower and Web Sites

With interest I read the column chronicling the history of the [Macmillan Digital] Web server ("Home Page," Feb. 1996). I'm amazed, not by the growth of your popularity, but by the expense you've gone to! I can understand the need for a T-1--that makes sense given the graphic richness of your site. But a multiprocessor Sparc 20 with 128MB of memory to handle only 200,000 hits a day? Gracious!

I run the Internet Business Pages (http://www.ibp.com/). Our server is handling over 120,000 hits a day, and all we have is a 60MHz Pentium with 48MB of memory on a 56kbps line. We, too, are upgrading to a T-1 shortly, as our network connection is maxed out. But the little Pentium will be just fine until we hit about 350,000 hits a day.

Why the big difference? I'm not sure, although the large number of image maps, CGI scripts and other inefficient use of your server may account for some of your need for power. A friend of mine operates a Web server that processes two million hits a day on a custom diskless 90MHz Pentium server with only 48MB of memory. That's up to 300 hits per second at peak, all on a single-processor Pentium. He gets a lot more work for a lot less power, because his software is good.

Maybe you need to examine your software configuration. Throwing more horsepower at your problems will probably solve them, but you can do much better for much less money by applying a little brainpower instead of only applying money for hardware.

The Web is a tremendously powerful tool that is regrettably misapplied in some cases and is responsible for most of the Internet traffic (see ftp://nic.merit.edu/-nsfnet/statistics/1995/nsf-9504.ports). Doing things right makes the tool more powerful; doing things wrong costs people time and money.

My bottom line: 200,000 hits a day is not much load for a well-configured Web server. Your article made it sound like you needed killer hardware to serve that many hits. Unless they are all CGI scripts (which they shouldn't be!), a processor with 25 percent of the power of your Sparc 20 MP is more than adequate to the task.

Daniel V. Klein
Pittsburgh, PA

I agree with many of your comments, but pure hits alone are not a measure of the load on a system. Our site is not as single-purpose as the sites you referenced. We don't just run a plain Web server. We also run WAIS (an FTP server) and Oracle, and most of the content users see is generated on demand by programs written in Perl and CGI. All the static pages are preprocessed as well for each hit, since they all have to have the shopping bag numbers inserted in links (and most need an include file added as well). If we only had static pages, didn't use any search engines, didn't have any include files or server-side executables and didn't have the shopping bag numbers, we probably could run our site on a Pentium-based machine. However, if we did any of those things, we also wouldn't be a world-class site.

--Jordan Gold

Off the Mark

An error appeared in your April issue ("Behind the News: Once Again, UNIX Unification Announced"). The article incorrectly states that NEC "has set up a facility in San Jose, CA, to develop a low-end Mips chip to run this new operating system on small-scale devices, such as point-of-sale systems."

What NEC announced in conjunction with the HP/SCO announcement is that we, in association with our Mips ABI partner companies, have formed the Golden Gate Operation as a division of NEC Systems Laboratory in San Jose. One of the purposes of this division is to work with HP and SCO to develop the next-generation UNIX operating system and to make it available on Mips RISC-based computer systems, ranging from workstations to massive SMP systems. These systems will use various MIPS chips. There was no announcement of any new chip design or any project related to point-of-sale systems.

Kevin Payne
NEC Systems Laboratory
San Jose, CA