The Passing of an Era

From CP/M to Windows NT

Chapter 1

Today, July 3, 1998 marks the passing of a computer era for me, and it is time to stop and document just what has come to pass in a methodological sort of a manner. I joined the computer world with sort of a bang in the spring of 1980 when in early January of that year, I placed an order with the local Radio Shack dealer for a newly released CP/M compatible system that Tandy Corporation had announced. It featured a Z80 CPU with one wait state, 64k RAM and a single double density 8" floppy drive integrated with a monitor in a single box.. Expansion capacity was provided for up to 3 additional drives each with 596k of storage.

That generation of hardware continued to serve me for nearly a decade. To be sure, the single sided drives became double-sided floppies and half high drives to boot with 1012k of capacity per floppy, and in the later years even hard drives appeared of the 5 megabyte variety, but the technology was essentially the same. For printers, I began with NEC spinwriters which cost in those days around $2,500, and then bought some NEC 860 laser printers for similar money. Indeed the first single user work station between the computer, extra drive and printer cost about $7,500.00 in 1980 dollars which was also about the price of a new full sized Oldsmobile.

In the mid 1980's I was introduced to MP/M II as a beta tester, and ultimately created a 4 user MPM II based system on a 4 floppy drive multi user system still grounded on the Z80 technology, but alas, the DOS world was passing me by. It was fairly easy to ignore the early PC's as they were because the Z80 system carefully configured would do as well and had more drive space with the generous 8" floppies. I began to see the hand writing on the wall, and in 1987 I acquired a NEC laptop based on the NEC clone of the Intel i8086 At 12 pounds with a 20 meg HD and a flat screen it was state of the art. I also used it to learn about DOS. Finally, Intel got it right with the i386 processor and in the spring of 1989, I ordered a full tower system from an emerging giant Gateway who featured upscale Dos systems and aggressive prices. $4500 got me a i386 with 4 megs of ram and an astounding 150 meg ESDI hard drive along with a NEC multisync monitor built to the then just released VGA standards.

The same winter Digital Research released Concurrent DOS 386 which implemented the ability of the i386 to run multiple virtual DOS environments as once. With a serial Port card I was able to slide right into the DOS world. I got some PC compatible WYSE terminals, and bought a DOS versions of Wordstar and smoothly slid into the DOS world with a multi-user system that was upwards compatible with my MP/M II system. I made up a Null modem and used a Telecom program to download critical files and went on.

Once I got the first 3 users or so going on the Multi-user Dos, I merely had to push the technology a bit. I went to 8 megs, then 16 megs of Ram, and the hard drive went from 150 megs to a 768 meg replacement. I still have that full high 5.25 inch form factor drive around here. All 10 lbs of it. It served me well, and backups would take scores of 5.25" floppies. Oh to be sure, I had a 3.5" drive in the system as well-those were the days of dual drives, but the 3.5" floppies were much more spendy so I used the larger ones. I also added more terminals, and by March of 1992, I finally retired the last component of the Z80 based computers. By then I had recompiled my accounting system which I had written for CPM in the early 1980's and brought it up on March 15, 1992, thus ending the first Epoch of computer history.

I stretched the multi-user system to the whole office making 7 workstations for 6 people, and gradually the 386 motherboard gave way to a 486-33 motherboard, and the RAM went to 32 megs and finally 64 megabytes. In December of 1993 a Pentium-90 retired the 486 and forever ended the performance complaints that sometimes dogged the multi-user system. With the Pentium motherboard came integrated IDE drive subsystems, and a 1 gig Hard drive. It cost $500 and the first one I bought was a lemon and they wouldn't take it back, but the second one still works. Somewhere along the line, I retired the Western Digital ESDI controller in favor of a hardware caching ESDI controller. There were two reasons for this. First, the model of controller that I had had a reputation for locking up from time to time during file reads when under heavy stress. Apparently a single user system couldn't stress it enough to cause the problem, but several people using the system could. Secondly, memory was always short and the hardware cache provided performance that I wanted, rendering unnecessary the tricks that I some times used of loading the print spooler into a virtual disk, or loading common program overlays into a virtual drive. The other thing I did, of course is to replace the 150 meg issue drive with a 700 meg one. I paid around a $1,000 for a factory refurbished drive--full high 5.25" monster. They will make a good door stop even on a windy day. The ESDI drives had to be low level formatted. To install one, you selected a type 1 drive of the choices in the BIOS. You then loaded debug and issued a GO command and jumped into an onboard BIOS extension in the drive. From there you could lock out the bad tracks, set the sector size and do a low level format. These drives came with a factory locked out bad track list in a little envelope glued to the outside of the drive. The drive I had, contained quite a few bad sectors, but I low level formatted with with a 'spare sector' per track. I lost about a 100 megabytes of capacity this way but eliminated all but two hard errors on the drive which spent their life locked out with be red 'X's when ever I did a high level surface check.

Also along with the Pentium came MaxStations which at long last meant that the WYSE terminals and their 38.4k connections gave way to regular VGA monitors and VGA graphics, and finally in the summer of 1995, a new version of multi-user Dos that would even allow Windows 3.1 to be loaded, but of course WIN95 killed off WIN 3.1 just as I was getting that capability. I bought a lot of Phillips Magnavox 14" monitors and boy were we happy. Now we had color. I moved to a Pentium Pro 200 and upped the RAM to 128 megabytes. Meanwhile CDROM drives and ZIP drives appeared, but neither worked well with Multi-user DOS now called REAL32. By now the NEC laser printers were forgotten and some very nice dual bin Hewlett Packard units were in service.

My upgrade to the Pentium Pro 200 was a trial balloon to see if it was practical to run multiple windows 3.1 sessions on the system.. What I learned was that while it was technically feasible, the horsepower required would even cause performance problems with a Pentium Pro 200. More over by 1997 it was apparent that the application software was quickly becoming WIN 95 only in nature. Additionally, pending personal changes at the office made it apparent that change was necessary as the incoming people were expecting that current technology would be in place. With Law libraries being on CD and new software demanding WIN95 or better, it was clear that I was at the end of the road. Just as I had in 1989, I need to jack up the dust on top of the computer system and slide a new system underneath. The rapidly declining prices of computer hardware made a bold stroke. To be sure, I could have replaced my old Max Stations with new SVGA ones to get much needed SVGA resolution and I could have fragmented the 7 user system into a couple of segments and carried on with the ultra thin networking design. I even contemplated some of the 'thin network' designs being discussed by Microsoft, but decided that I couldn't wait. I had to move now-now being the spring of 1998.

With well loaded full fledged computers under $900, It was my judgment there was no longer any space for 'thin computing' I made an executive decision to go with a TC/IP network using a classic server and client combination. It was fairly easy to conclude that it should be based on WINDOWS NT instead of either a UNIX box or Netware. My logic was simply that WIN-NT works, while UNIX boxes are expensive and Netware seems to be fading. While I would have preferred to do it with NT-5, that product was deferred, beyond my window of time so NT 4.0 it was. I did hold the project for WIN 98 for the clients as it did become available within my time frame.

My specifications called for 7 or 8 workstations, a file server, and a couple of printers. The target application software package was to be Corel Office 8. However, during the transition period current Dos based software packages would be moved over, including Wordstar, and Paradox for Dos. Migration to the Corel Office was the easy way to go because though fading compared to MicroSoft Office, it is popular in the legal profession and will simplify the migration from Paradox for Dos, as Paradox for Windows is file format compatible with the Dos version and will work interactively with the DOS version. While it is not clear how much longer Corel Office will survive the MicroSoft office onslaught, I was my judgment that the Paradox compatibility and experience of incoming personal with Corel Perfect justified going that direction even though a side step to MSWord may be required in the foreseeable future.

Given that my existing system used VGA monitors, my transition plan consisted of sticking a network under my existing monitors and copying over major application software and databases. To this end, I acquired a Gateway Pentium II - 350 with a 10 gig drive for a file server, and had 7 client boxes built to order locally. They are P-233 MNX systems with 64 megs Ram, a 4.3 Gig HD, and a 100x CD drive and an internal Zip drive. There were clearly low end systems by today's standards tweaked with extra memory, and a nice assortment of drives. The most costly decision was the Zip drive. On a system that cost $825, it was $99 of that cost but I felt it was justified. I didn't want to spring for a more expensive solution and I felt strongly that some credible removable media was needed. The standard 1.44k floppy is a joke in the days of multi gig hard drives. Clearly the market is ready to explode to some sort of replacement for the classic floppy. The ZIP disk is expensive and only an interim solution, but the magic solution hasn't happened yet.

I got these systems with WIN 98, and went straight away to FAT 32, and installed my Netgear FA 310tx Fast Ethernet PCI network cards from Bay Networks. Though apparently a good card, they proved to be a royal pain to implement because they are not cards that are on the 'drivers included' list of either Win 95 or Win 98 though I was to learn later than WIN98 lacks current drivers for most network cards.. Consequently it was necessary to stuff an included floppy disk at the system during the install routine which would not have been necessary if I had selected a card that Win98 could correctly recognize. I selected this card because it was cheap ($29.00) and supported Fast Ethernet (which I wanted). For what it was worth, my decision for Fast Ethernet was against almost all advice that I received on the subject. Classic Ethernet (10 times slower) can be configured to work properly in even very large configurations so what is the point, I was told. They may be right, but the extra cost was minimal so I went for it, on the theory that faster is better. Currently, most PCI based Ethernet cards are available at modest cost with auto-speed sensing. The difference is in the hubs. The cabling is the same, or at least should be. Fast Ethernet (100base tx) requires a category 5 twisted pair cable which is suitable for 10base t as well. While there is legacy wiring of a lower standard for 10base-t there, is no point in installing such wiring now.

What I learned about network design is that you break the network up into subnets (typically workgroups) as load and distance demands and interconnect the subnets with switches. Within the subnet you use hubs. Hubs are not very smart and are much less expensive than switches. Switches can be used to change speeds from classic to Fast Ethernet, or to effectively provide a dedicated circuit on the net. The advantage of Fast Ethernet is that you can shift the balance to more hubs and fewer switches. Of course, for my small system, I have a couple of hubs daisy chained and no switches at all. The downside of the Fast Ethernet is that at this writing, network cards for printers of the Fast Ethernet variety remain expensive and hard to get. If you must drive your printer off the net (as opposed to through a PC) you have to fork out extra bucks for either an internal or external printer interface card. While is suspect that historically, printing though a PC created performance issues for that PC, today's PC's seem to have the capacity, but of course the PC working as a printer server must be 'on line'. At least initially, I elected to put an EIO card in my HP 4000 series printer, and with the Jet Admin software you can display on the server even how much paper is in the paper drawers. Older HP printers take the MIO card which is more expensive, and I deferred that, using a PC and parallel port instead.

Once I made the decision to go with Fast Ethernet, I was ready for a trip to Fry's to get on with it. My first go around there produced a Netgear FE108 hub which is a low cost unmanaged hub with one daisy chain port. What this means for the uninitiated is that the hub actually has nine RJ-45 ports, but you can only use 8 of them. One port is a cross jacked version of its clone. If you are going to hook a couple of hubs together, you plug one end of the cable which interconnects the two hubs in the 'daisy chain' port and the other end in a regular port (on the other hub). If you do this you can use a standard 'patch cable' which is connected straight through. Otherwise, you must use a 'cross over cable', which is sort of the networking version of a 'null modem' used in the serial port days. Some hubs are called 'stackable' and have a short high capacity proprietary interface cable to interconnect them but that was not my priority. I was going to need two hubs in different locations as my office was physically spread out quite a bit and this way it would save a lot of wire. In the end, I located one hub near the server (just a few feet away), and the second hub daisy chained on a 100 foot cable with 5 or 6 ports in use on each 8 port hub. My second hub purchase was a mistake. I bought a 'dual speed hub' thinking I was getting something. To be sure, I could use it but I had confused it with a switch believing that it could convert from one speed to the other. Alas It would work at either speed, and any port would work at either speed, but the high speed and classic speed ports did not communicate with each other so it was really just a hub that could be used at either speed and without a switch was not going to provide what I thought I was getting-- a hub that would support both speeds of Ethernet, and magically convert one to another. You see, I was interested in some classic Ethernet support because I could get classic speed ethernet cards for my older Hewlett Packard printer on the cheap, but Fast Ethernet cards of the MIO style printers cost a fortune (nearly $400 each). However, If I every buy a dual speed switch, I can split the hub and run part of the hub at each speed. Beyond that, the wiring for a network is no more complicated than installing a phone system. Buy the cables and plug them in. Raceways, of course, can be a problem, but if you are short on imagination at where to string the wires, it is always worthwhile to observe on how the phone guy got the phone wires in. Usually the computers are going to be on desks next to phones, and if the phone cable got there, there must be a way. The preferred method would be to buy a roll of wire and a crimper and make your own cables on the fly. This assures that they are the right length, and also saves money. However since I had never crimped my own before, I used more expensive pre made cables with come in two lengths-too long and too short. It may seem juvenile, but it makes sense to draw a diagram of what the wiring is suppose to look like before you start, and contemplate appropriate locations for hubs as well. Bear in mind that the hubs (and cables) are security issues, because if James Bond sticks a cable in a blank port on your hub (or splices his own hub into one of your wire runs) Mr. Bond will then have a terminal on your system which you don't know exists.

Once the wiring is in hand it is time to install the server. The server does not have to be anything special. With due consideration to the fact that it will be a shared facility, any computer that will run your server operating system will do. Larger networks often use a Unix based mini computer, but for a small network, I selected Windows NT. In fairness to Windows NT, it is also very scalable to larger networks, but in my design, I simply bought a 10 user version of Windows NT, and modestly upscale Gateway computer. You get what you pay for in servers, and what I got was modest consisting of a Pentium 350 with 64 megabytes of Ram and a 10 Gigabyte IDE style hard drive. I could have designed in more performance and reliability by using a SCSI drive system, with or without a Redundant Array of Inexpensive Devices (RAID features). SCSI drives are generally regarded as providing better performance in the multi-tasking context of a network, and have historically been available in larger capacities. My bow to reliability was a battery backup for the server. I got an upscale on that logs the power quality to the server via a serial port, and the in the first week I learned that voltage sags sufficient to kick the back up on were daily occurrences. This sort of confirms what I have suspected for years-namely that many of my computer crashes over the years are likely power quality related. I've been running computers for 18 years but never before have I used anything except a fairly low cost surge protector. For that it is worth my engineer friends tell me that a $10 surge suppressor is worth about what you pay for it, and that even that effectiveness diminishes over time so old and particularly cheap old surge suppressors are best used for door stops.

Installing Windows NT

At some point you have to face up to Installing Windows NT on your server. In some ways the install is like any other. You belly up to the computer with 3 floppies and a CDROM. And straight away, you have to start making decisions which are going to shape the whole network with little warning. The first problem is deciding on a format for the hard drive. As with DOS you can run FDISK and partition the hard drive as you wish-almost. Windows NT supports the classic FAT 16 file format (used from about DOS 3.0 to Windows 95) and an NT unique format called NTFS. NT 4.0 Does not support the FAT32 common to WIN98 and later editions of WIN 95 that allow partitions larger than 2 gigabytes. For lack of a better plan, it is customary to create a modest sized boot partition <less than 2 gigs> in FAT 16, and then format the rest of the system in NTFS. The file sharing routines of FAT 16 (and FAT32) leave a lot to be desired so figure on the FAT16 area to be the portion of the hard drive used locally on your server for server use, and do the 'serving' on the NTFS partition. The latter provides security features, stability, and proper sharing none of which is present in the FAT16 model, but having a FAT16 boot partition is critical if you are to boot the system to any other operating system.

Once you get past the file formatting you must face declaring one or more network protocols. Several are supported, but use of multiple protocols will add system overhead. Clearly the flexibility is needed so you can match another system if you are going to be connecting. TCP/IP is the protocol of the internet, and my choice for that reason since I had no legacy network with which I needed to integrate. In a TCP/IP system every device on the network that is anything is identified by an IP address. In the old days IP addresses were passed out by the system administrator and hand coded into every workstation and printer-a real pain. Fortunately, NT 4.0 support a dynamic addressing system called DCHP. This means that you give your server an IP address which you can select out of the sky if the server is not going to be integrated as part of the internet, and the server will dynamically assign IP addresses for all clients on 'as needed' basis meaning that you never have to think about them again. Once this is done is is sort of down hill. You will want to identify a 'domain' which is the big tent of which your users will be a member. This will get a rough installation made. Logically at a minimum you will have to create some users for the system establish some sharing rights for those users. In my case I made the entire 'second drive' with the NTHS format a 'shared drive' which would be available to all users. In turn that 'share right' shows up as an icon in the 'network neighborhood' section on the client systems. There it can be declared as a logical drive on the client and transparently integrates into the client system as a logical drive just as if it were physically located within the client computer. Once you get all the boxes checked properly on the client system, a log on to the client computer is also a log on to the network, and clears access to the shared facilities.

Printers are logically just folders, and are managed in the same way. If the printer is plugged into a client computer (typically via the parallel port) you install the printer to that client just as you would in the single user world. Then you 'right click' on the printer icon and set up the sharing which allows the others to 'see' the printer, and install it for use just as they would a local printer. If you have a printer with a network card which allows it to be connected directly to the network there is one more step. In the case of Hewlett Packard printers you run their "Jet Administration" software which locates it on the network and displays it on the 'where is it list' along with devices that might be connected to serial and parallel ports. The computer form which you ran the 'Jet Admin' becomes the print server, but when you are done with the installation, the network printer as with all the installed printers becomes just another icon in the 'Printer' folder which can be selected for use. As with any other printer, some printer will likely be designated as the 'windows default' printer with the other printers becoming ones that you can optionally select in most windows applications as part of the print dialogue.

To printer with DOS applications you will need to dig around in the printer Properties and 'capture' the desired printer as logical LPT1 or some other classic DOS output port if your application is configured to print to it. DOS applications that have print dialogue options to let you easily choose whether the printer output should go to LPT1 or LPT2 for COM 1 or 2 can be easily used to choose network printers. You simply 'capture' the various network printers under these designations and route the DOS print output appropriately. The only complication is that you must remember which printer you associated with which DOS port.

Chapter 2

The New Era Opens

I never did get around to say it specifically in Chapter 1, but July 3, 1998 was plug pulling day for me. I pulled the plug on my Multi-user Dos/Real 32 system (almost) and brought seven Windows 98 workstations up on a Windows NT network. I jacked up my old DOS programs including Paradox 4.5, Wordstar 7.0, Supercalc and a few others and moved them wholesale to the NT server. Ditto for all of my databases and directory structures. With the benefit of a few zip disks the transfer was pretty easy.

Since my old system was already using VGA monitors, they didn't change, and other than the addition of a mini-tower in every office things really didn't look much different to the users. Although most of my folks had no experience with Windows 95/98, the menu bar across the bottom, and the multitasking was enough like the multi-tasking supported by Real 32 that barely a glitch occurred in the initial transition. They were looking at the same monitor running the same applications. The only real failure over the near term was my inability to get my old self written CBASIC accounting program to run right. It always tested out find, but when the bookkeeper tried to use it, it would die.

Ultimately, I kept the Station.0 of the Rreal 32 system going for an extra month while I studied this apparent compatibility failure. I was continually getting error messages that it was unable to open its files. These messages always appeared when the bookkeeper would try to use the program, but not when I did. Ultimately I figured out that her problem was that she was trying to load and run other DOS programs at the same time, which I didn't do when I was testing it.

I circled the problem for more than a month before I solved it. My problem was that I was not creating a CONFIG.SYS file in the drive C: root and putting a FILES=60 or some other large number in it. Simply stated the default FILES value for the DOS box of WIN 98 was too low to run multiple DOS programs. Once I got it though my skull that I needed to create a traditional CONFIG.SYS file such as we did in the DOS days, for the DOS based programs that I was running on WIN 98, all was well. Real 32 was finally retired.

This did not mean that other hardware problems did not haunt me a bit. There were the monitors. I had 3 seventeen inch monitors and 4 older 14 inch monitors. The 14's quickly demonstrated that they were too small (and too burned in) to be satisfactory. Windows 98 displayed even the same applications is different proportions and the burn in tracks from the previous life just created blurs. Of the 17s, one inexplicably emitted a cloud of smoke and went dark, any another proved burned so dark as to be unreadable with a graphical display.

The bottom line on the monitors was, that except for one that I had replaced just a few months previous, all the old monitors soon went to the dumpster. The good news was that several of them hung on as the price dropped over the summer and fall of 1998 and the last couple change outs were for less than $200 dollars each for seventeen inch monitors. The essentially failed effort to use the old monitors saved me some serious dollars just in price drops.

I reluctantly provided 15" monitors to a couple of people who insisted that 17" monitors were just 'too big' for their desks. Everything else went 17".

Although every computer has a sound card in it, I have been spotty about configuring for sound as I write this in mid November of 1998, only 3 of the 7 stations have working sound systems. No special reason, except just haven't gotten to it, and there doesn't seem to be a need. Of course along the way, I got to deal with a few other hardware issues as well. My locally produced inexpensive computer boxes were equipped with Samsung hard drives. Life has been much easier since I took some remedial action there. One made its failure diagnosis easy, one day it just stopped spinning. I didn't loose anything of importance because I am using a thin client configuration implying that everything important is on the server Hard drive but it was an abrupt end which caused me to focus on the hard drive reliability issue.

On another system, about once a week, vast contents of the hard drive would simply disappear. The disappearing files would normally be the ones that I was using (like all of the important files in the Windows directory). After reinstalling Windows 7 or 8 times over a couple of months, I decided it wasn't just an unlucky installation. I chucked the hard drive and while I was at it, I noticed that the CPU fan was dysfunctional. and I replaced it as well. Since I changed a couple of things as once I don't know which one was the culprit, and don't much care, but the system now works, and I am a very happy camper. It has only a 1 gig 'take out' drive from another system that I upgraded, but so what-that is enough for Win 98 which is all I really need.

One of the nice things about fast ethernet is that it is fast enough to make the network drives very transparent. There is little notice that the network drive is 'slower' that the local drive. In the 'real world' the response time is close enough to the same that no one notices.

A particular high point so far has been the server reliability. In the 4 .5 months since it came on line there have been no unscheduled reboots. The UPS carried it right through one momentary power outage. I do not as yet have UPS coverage for my hubs or workstations so the power outage kicked all of them down, but when they came back up the server was waiting. I am quite convinced however that UPS protection for the workstations and hubs is worthwhile and it is a priority on my things to be done list. The momentary blink cost about an hour of lost production and when you figure that those happen a couple of times a year, (where my system is) the battery pak looks attractive, and will be in a budget soon.

Software migration

The perceived need to migrate to Windows software was, of course, the driving force for the whole transition to a Windows based network. That migration has not met with as much user acceptance as I had hoped. I ended up waffling on whether to go to MSWORD or Corel Perfect. At the moment I am headed toward the latter because it does a better import of Wordstar files than MSWORD does, particularly since I haven't successfully installed the MSWORD/Wordstar translator. Most of us still prefer to use Wordstar, which I guess is OK. It runs grandly under Windows NT.

My plans to migrated to Corel Paradox for windows which is file compatible with the old Borland Dos version has been completely stymied by my own inadequacies. Paradox for Windows does not support a macro language (unlike the Dos version) and use of the OOPS interface that is provided has so far proved to be beyond my technical skills. Consequently Paradox for Dos it will be until either a version of Paradox supports macros or until I figure out OOPS, or I change database programs to one that I can program. The good news is that Paradox for Dos is stable even under NT so there is no real pressure to change.

In other areas I have been able to move forward. One of my early priorities was a calendaring program. On DOS we were in a time warp stuck with Borland Sidekick 2.0. To be sure, it was network aware and provided the interactive multi-user interface that I wanted, but it was very weak in providing a printouts. I evaluated new sidekick programs and found them wanting in that they were not network interactive. I circled MS Outlook and it seemed to expect that I would have MS Backoffice in order to be network interactive. I also circled Corel Central and found that it would do everything except work. Finally, I discovered Organizer from Lotus. I first installed version 97, but am in the process of upgrading it to version 5. It allows multiple people to view/change a single calendar file, and in fact display multiple files as differing 'sections' of the same calendar which is just exactly what I wanted. Likewise it supports a variety of prints so that transition has been made and we are all happy about it.

New things that have become possible include the utilization for the first time of law office specific software for legal research as well as specialty software for document preparation.

Places to Go

I've had some failures as well. The most spectacular has been with my attempt to integrate faxes into the network. I undertook that by implementing and installing Symantec Faxworks 9.0 which is suppose to do exactly what I wanted. It is a TCP/IP based product that is suppose to answer the faxmodem, receive the faxes, print them out of a designated printer and make the available for forwarding to other workstations as well as permit the origination of faxes anywhere on the network

Well, I said I never had had to make an unscheduled reboot on the server. But I had to schedule a few because of Faxworks, because it would lock up the workstations, and gum up the network beyond all belief. After about of month of greif with it, I removed it from the system and no one has asked for it back. A functional fax manager is still awaiting installation and use on my system as I write this.

Another semi failure has been internet access. My design calls for a secure shared access which will allow any user to gain internet access via a common dialup modem. This implies the use of such things as a router, a proxy server, or a fire wall or some combination of the above. I have not gotten there from here and to date internet access consists of a couple of dial up modems in workstations which are hooked to different outside phone lines. This approach while apparently not uncommon, is not 'networky' and provides an opportunity for improvement-one that will be necessary particularly if the Internet is to be further integrated into the system or a highspeed internet connection is to be considered. If I understood these issues better, I might get there from here more quickly.

The Path

Beyond things that I have tried and failed at, there remain further places to go and people to meet. Dial in access for the work at home crowd is a possibility which I have not explored.. I simply have not been there. Other things which are pretty high on the list include adding a network class scanner and a quality color printer. The legal practice is wide enough now to require the preparation of presentations. Flashy color things are a necessity. I've thought about hanging this stuff on the server that I have, but have thought better of it. My present plan is to add another computer to the network which I may configure as a backup server. Ultimately, I see this piece of iron supporting the failed 'fax server', a scanner, and a color printer.. To be a backup server, it would need NT on it, which is the rub. My instinct is to buy a USB (Universal Serial Bus) scanner, but NT 4.0 hasn't gotton around to support the USB just yet, unless that support is in the just released 'Service Pak 4'. What I really need is NT 5.0 which I just heard will be known as "WIN 2K", which will have all the Win 98 support features, however, I am getting the feeling at the moment that WIN 2K may not be here before 2K and I want to move before that.

The price trends for scanners and color printers have been very buyer friendly so my inclination to make this a project for 1999 seems prudent at the moment.

The Hub Error

The final thing that I need to address is the Network Hub issue. The first hub I bought to prototype up my system was a NETGEAR FE108 hub. It is Fast Ethernet only and of the lowcost non stackable variety. As I completed my prototyping of the network, I became aware that some things were very expensive to do in Fast Ethernet (Printers and routers) to be exact. Since my physical configuration called for a second hub anyway I purchased a Dlink dual speed hub believing that I had solved my problem. Only after I bought it, did I realize that this model from Dlink was literally just that a dual speed hub. You could plug both classic ethernet or fast ethernet into any of the 8 ports as you wished, however the hub would by design partition itself and the fast would only talk to the fast and the classic would only talk to the classic. You either had to use a separate switch to get the classic and the fast to speak to each other, or stack this hub with a special and very expensive switching version which contained an internal switch to connect the fast and classic partitions.

As the year has worn on, however, this problem has been resolved by technological advances and price reductions. Low end network hardware has advanced capabilities and gone down in cost. Dual speed hubs with switching capabilities between the speed sections are now readily available, from several manufacturers and the Netgear version, a DS108 which is an unmanaged, non stackable dual speed 8 port model (one like I need) is priced for less than I paid for the single speed hub 6 months ago.

I can dump my present hubs and replace them with switching dual speed hubs which will permit the use of legacy ethernet cards in the older HP printers and come out ahead of the cost of buying fast ethernet cards for my older HP printers. Presently I have both of my older HP laserjets printing on parallel ports of workstations (which are shared). This works, but requires those workstations to be running before the printers is accessible and implies that the workstations are also doubling as print servers. They have the horse power to do it, but it is not as tidy as having the printers directly on the network.

I'm a little embarrassed to dump hubs that are only a few months old, but the kind I wanted simply did not exist a few months ago. To be sure I could have going to a switched network and gotten there from here, but a switched network is much more spendy that a 'hubbed' network, and the total size of my network (less than 16 connections) makes the implementation of a switched network seem like overkill.

Anyhow, besides UPS facilities for the workstations, a Scanner, and a Color printer, a couple of the newer dual speed switching hubs are on the Christmas list for upgrades.

- - Updated 12/16/2012
- - Updated 04/25/2008
- - Updated
- - Updated 2/8/01
- - Updated 03/21/2008