TTG Home Robert Home Daynotes Journal Home Journal for Week of 20 August 2001

photo-rbt.jpg (2942 bytes)Daynotes Journal

Week of 20 August 2001

Latest Update: Friday, 05 July 2002 09:16
 

Search Site [tips]


Click Here to Subscribe

Visit Barbara's Journal Page

Monday, 20 August 2001

[Last Week] [Monday] [Tuesday] [Wednesday] [Thursday] [Friday] [Saturday] [Sunday] [Next Week]
[Daynotes Journal Messageboard]  [HardwareGuys.com Messageboard]

08:48 - Barbara is home! She arrived about 10:15 last night. I joined the dogs in a barking celebration and circle-dance. The dogs are pleased because Barbara will remember to feed them and play with them. No one got in trouble for the broken porcelain bird. Barbara just passed it off with the comment, "These things happen." She's off to get her hair cut this morning, and then we'll do all the normal Sunday stuff--cleaning house, doing laundry and so on. I may take a break today. I worked all weekend.

Does anyone know if Windows 2000 Professional Backup delivers warning prompts when it's time to clean a tape drive? I used to run a DDS-3 tape drive under NT4 with Arcada/Seagate/Veritas BackupExec. It would prompt me periodically that it was time to clean the tape drive. I'd always assumed that that was a function of the tape drive rather than the software, but perhaps not. As I was running the backup yesterday, it occurred to me that I hadn't been prompted lately to clean the drive. Doing it is no problem. One just sticks a cleaning tape in, waits a few seconds, and the tape ejects automatically. The drive recognizes the cleaning tape as such, and automatically does the required cleaning pass.

I'd assumed that running the cleaning tape reset some sort of counter in the drive itself, and when that time expired the drive sent a warning message to the software. But either it's purely a software function, or Windows 2000 Professional Backup isn't bothering to deliver the messages. I've cycled through half a dozen tapes over the course of the month or so, backing up 15 GB or more to each tape. Surely it must be time to clean the drive?

The meter on Barbara's old mechanical Pentax MX camera failed while she was on her trip. It had been giving signs of imminent failure, but now it appears that the meter is gone for good. We still have other Pentax K-mount bodies, so that's not a real problem, but Barbara now says she wants me to buy her a digital camera of her own. So I guess I'll start looking at alternatives.

09:47 - Thanks to Roland Dobbins for the pointer to Redmond Linux. As I've said repeatedly, Linux isn't going to become a mainstream desktop OS until it looks, feels, and works like Windows. It needs to play nice with Windows networking, present volumes, directories, and shares in the standard Windows way, and so on. None of this /dev/rfd0 stuff. If a user wants to access the floppy drive, he expects it to be A:, not /dev/rfd0. Little problems like that are showstoppers for most users, and one of the two main reasons why Linux has essentially zero penetration on the desktop outside the enthusiast community. The other, of course, is applications. Users may not insist that IE, Microsoft Office, and their other main applications run on Linux, but they do insist that what applications they are to use on Linux be compatible at the file level with mainstream Windows applications.

It seems that the folks at Redmond Linux are creating the first Linux distribution that recognizes these indisputable facts. Hundreds of millions of people already more-or-less understand how Windows works. They're not going to change their way of doing things, so if Linux is to succeed on the desktop, Linux must adapt itself to the expectations of those hundreds of millions of users.

From looking at their web site, I think Redmond Linux has a lot of potential. They're in beta right now, with the shipping product expected in about a month. I haven't downloaded the beta, both because I don't have time to play with it right now and because with the release version imminent it makes sense to wait for it. But you can bet that in a month or so I'll download the release version and play with it. If it's as good as I hope it will be, I'll pay for the product. They even accept PayPal.

Click here to read or post responses to this week's journal entries

Click here to read or post responses to the Linux Chronicles Forum

[Top]

Tuesday, 21 August 2001

[Last Week] [Monday] [Tuesday] [Wednesday] [Thursday] [Friday] [Saturday] [Sunday] [Next Week]
[Daynotes Journal Messageboard]  [HardwareGuys.com Messageboard]

09:32 - News reports this morning say that Excite@Home is about to go belly-up. S&P downgraded its shares from B- to CCC, an indication that S&P doesn't think the company is long for this world. It appears this company is another victim of the deranged thinking that ISPs can somehow tie content and delivery mechanism. Superficially, it would seem that AOL validates that idea, but in fact most AOL users subscribe to the service because of its perceived simplicity or because it was the best or cheapest way to get Internet access. AOL users by and large don't subscribe to AOL because they want AOL access. They subscribe because they want Internet access.

I subscribe to Time-Warner's Roadrunner service because I want reliable high-speed Internet access, not because I care about any content that TW may provide. As far as I'm concerned, TW content is just another resource on the Internet, one I don't care about. Attempting to tie content and delivery makes about as much sense as trying to tie content to telephone service. Or water, electrical, or natural gas service, come to that. Internet access is a utility, pure and simple, and the companies that realize that will be the ones that succeed.

AMD seems to be hemorrhaging customers. I mentioned last week that IBM had dropped AMD from their product line. That actually happened back in May, but no one noticed until recently. Now The Register reports that MicronPC, the third largest US OEM, will no longer sell AMD-based systems to small- and medium-size businesses or the government. Tiny, a large UK OEM, has both dropped AMD processors from their product line completely. AMD may soon find themselves in the same position Apple is in, with their fastest systems running at half the clock speed of the fastest Intel systems. Argue actual performance all you want, but the simple fact is that consumers, including corporate ones, buy clock speed rather than performance. Things are not looking good for AMD at this point.

The Register reports that Taiwanese DRAM makers are prepared to stop shipping memory if prices fall much further. Their drop-dead price is apparently about $1.35 per 128 Mb part. That translates to a price for a 128 MB DIMM in the range of $15. I don't believe them, though. A memory fab is an expensive piece of real estate, and the actual cost of producing a memory chip is relatively small. If you're paying the mortgage on a memory fab, you'd rather produce chips and get at least some income from them than have the fab sitting dark and producing no revenue at all. Still, at $1.35 per chip, we must be nearing the point where the cost of raw materials, electricity, and labor is approaching the selling price. I'm no expert on the economics of chip production, and I'm sure it varies from plant to plant, but I suspect that the DRAM producers won't really shut down their plants until the price per chip approaches $0.50. At that point, you'd need only $4.00 worth of chips to make a 128 MB DIMM.

The forecast for tonight is for clear skies and cool temperatures. Assuming that holds up, Barbara and I will probably take the telescopes out tonight to see what we can see.

Click here to read or post responses to this week's journal entries

Click here to read or post responses to the Linux Chronicles Forum

[Top]

Wednesday, 22 August 2001

[Last Week] [Monday] [Tuesday] [Wednesday] [Thursday] [Friday] [Saturday] [Sunday] [Next Week]
[Daynotes Journal Messageboard]  [HardwareGuys.com Messageboard]

09:39 - As Mick Jagger said, you don't always want what you get, or something like that. Here's a fascinating item on eBay, a copy of Book 1 of Amateur Telescope Making, the classic work on the subject. As the description says and the photographs confirm, it was printed in 1966 and signed by the editor, Albert G. Ingalls. Not just signed, in fact, but inscribed, "To Larry: Thanks for your great help in the making of this work. I couldn't have done it without you. Al Ingalls". So what's wrong with that? Albert G. Ingalls died on August 13, 1958.

We did end up going up to Bullington last night. The moon was only about three days old and set early, so we decided to look for faint fuzzies and other stuff that's difficult to find except under dark skies. We ended up bagging twenty or so Messier Objects, Uranus, and Neptune. The latter two were visible only as teeny, tiny bluish discs, but they did show a disc so they were clearly not stars. I also saw Pluto, but I'm not counting it because I'm not sure which of the myriad tiny points it was. At magnitude 13.8, it's just within the capabilities of our scope. But I did have it in the field of view. Somewhere.

The days are definitely getting shorter. Sunset was at 20:04. We got to Bullington by about 20:15 and got set up. It was dark enough to see things by 21:00, and we stuck around until about 23:00. By that time, our Telrad had dewed up, our finderscope had dewed up, and our eyepieces were beginning to dew up. All of the charts I'd printed were soft and soggy as well. But it was a successful evening. I'm looking forward to evenings this Fall and Winter. The weather will be cool, the humidity will be low, and sunset will come early. We'll be able to make an early dinner for my mother around 16:30, head off for several hours' observing, and still be home by midnight.

Back to work on the Motherboards chapter. I have that as well as the Processors and Memory chapters in progress. Motherboards is turning into a monster and Processors already is a monster. But I need to expand coverage of those and other subjects, because PC Hardware in a Nutshell is O'Reilly's only hardware book. Originally, PCHIAN was going to be the small, quick-reference version, and the "big book" Pournelle and I were working on was to be the huge compendium that would compete directly with Scott Mueller's Upgrading and Repairing PCs

But since O'Reilly cancelled Pournelle's and my big book, I have to make PCHIAN stand alone. O'Reilly won't publish a 1,600 or 1,800 page monster, but they will probably let me get away with an 800 to 900 page Nutshell. So my job for the next edition is to cram more useful information than those doorstop Frankenbooks contain into half the space. I think I'm doing that, but we'll see. As I described it to my editor, I'm trying to make PCHIAN into a pocket battleship. Smaller than the big guys, but packing an equal wallop. In the interim, don't hesitate to buy the current edition. The next edition will be awhile in coming.

Click here to read or post responses to this week's journal entries

Click here to read or post responses to the Linux Chronicles Forum

[Top]

Thursday, 23 August 2001

[Last Week] [Monday] [Tuesday] [Wednesday] [Thursday] [Friday] [Saturday] [Sunday] [Next Week]
[Daynotes Journal Messageboard]  [HardwareGuys.com Messageboard]

09:09 - More bad news for Gateway. The Inquirer reports that S&P has downgraded Gateway to "junk bond" status. This after Gateway last week reported a $9 million loss on $1.5 billion in Q2 sales and the announcement that Gateway would depart the European market. None of this bodes well for Gateway, which is a shame. I remember in years past when Gateway was the bargain alternative to Dell. Gateway used similar or identical components, but charged a couple hundred bucks less than Dell did for a system. That price differential disappeared long ago, leaving Gateway attempting to compete head-to-head with Dell. Obviously, it's not doing that very successfully.

In other news, Intel's rush to the Pentium 4 continues. Here, supposedly, are the prices at which the Pentium 4 will sell as of next Monday:

1.8 GHz - $ 275
1.7 GHz - $ 209
1.6 GHz - $ 179
1.5 GHz - $ 148
1.4 GHz - $ 140
1.3 GHz - $ 132 

If these prices are accurate, and I suspect they are, Intel is really putting the screws to AMD. AMD cut its prices Tuesday in expectation of Intel's Sunday cuts, but I suspect AMD's price cuts aren't enough to move their processors against these Intel prices. Clock-for-clock, Intel processors sell at a premium over AMD processors, which suggests that AMD would have to price their 1.4 GHz Athlon at something less than $140. Instead, their new price list puts the Athlon/1.4 at $253, the Athlon/1.33 and Athlon/1.3 at $230, the Athlon/1.2 at $199, the Athlon/1.13 and Athlon/1.1 at $179, and the Athlon/1.0 at $160. I don't think AMD is going to move many processors at those prices. And, of course, Intel will soon release a Pentium 4 running at the magic 2.0 GHz, putting AMD further behind the curve.

The actual performance of the two processors is another matter, of course, but, as I keep saying, people don't buy performance, they buy clock speed. I suggested to AMD more than a year ago that they should use the old Cyrix Performance Rating (PR) method, labeling their processors with a number that reflected their actual performance versus the Pentium 4 rather than the actual clock speed. My guess is that AMD may at some point become desperate enough to use this hokey dodge as the clock speed differential between their processors and Intel processors continues to widen.

On a related matter, I commented here some time ago that I expected Intel to ship their i845D DDR-capable chipset sooner than the announced Q1/2002 ship date. The rumors are now widespread that Intel will in fact do just that to counter the VIA P4 DDR chipset. I expect to see i845D-based motherboards begin to ship in the next couple of months. I have no hard information to back up that suspicion, just some private comments from people who are in a position to know. There is also speculation that the 845D will be "crippled" to reduce DDR-SDRAM performance so as not to make Rambus RDRAM look bad. No one I've talked to seems to think that's likely to happen. Intel still has somewhat of a split personality about Rambus, but the trend is toward Intel becoming memory-agnostic. Intel very much wants the P4 to succeed, so my guess is that the 845D will be as fast as they can possibly make it.

Back to work on the book. I'm in the midst of "hard" chapters, so things are going slowly, but they are going.

13:45 - One good reason for upgrading machines rather than buying new ones is seldom mentioned. Microsoft OS and application software licensing policies. Many people don't realize that Microsoft OEM licenses are tied to the particular machine. For example, say 18 months ago you bought a Dell Pentium III/450 system that came with Windows 98SE and Office 2000. You might think you can buy a new system without software, retire the old system, install those copies of Windows 98SE and Office 2000 on the new system, and be perfectly legal. You can't. That software is tied by the license to the old machine.

This may not be a major issue for someone who has one computer, but what about the business that has 10, 100, or 1,000 of those Dell systems? If that business buys new systems to replace some or all of those old systems, it has to pay Microsoft (again) for the same software--both the OS and the applications. What's worse is that that "new" software is a new license, and Microsoft's licensing terms just keep getting more and more Draconian. But that doesn't mean you have no choice. Under the OEM license, you can certainly upgrade your systems by substituting a faster processor, more memory, a larger hard drive, and so on. In fact, you can do all of those. I am not a lawyer, but it seems to me that as long as the machine you're running the software on is still the "same" machine, you're legal. So what constitutes the "same" machine? Presumably that little serial number sticker on the back of the case.

So, rather than replacing systems wholesale, it's quite possible for a company to upgrade systems in-place. In the case of the company with 1,000 systems, for example, they may be on a three-year replacement schedule. That means they have to upgrade roughly 30 systems a month. If I were that company, I'd seriously consider rebuilding systems routinely rather than replacing them. Although it may seem that that would be a lot more work, the fact is that it wouldn't be. Very few companies simply order in Dell or Gateway PCs and plop them on people's desks. They go through the IS department first, where they're configured to the standard setup of that company. A good PC technician should be able to do a standard hardware rebuild--cleaning out the old system and replacing motherboard, processor, memory, and drives--on at least six systems a day, which means that doing 30 systems a month would require only one week a month of that technician's time. That is little or no more time than would be required to receive, unbox, setup, and test 30 new systems a month.

Standardizing upgrades would be little problem in most corporations, which tend to buy systems in large groups with similar or identical hardware configurations. A mythical average 1,000 PC corporation right now, for example, might have 100 Gateway Pentium/200 systems, 200 Gateway Pentium II/450 systems, 300 Dell Pentium III/550 systems, 200 Dell Pentium III/800 systems, and 200 Dell Pentium III/933 systems. Devising and implementing a detailed rolling upgrade plan for those systems should be straight-forward, and should cost substantially less for materials and labor than the cost of replacing the systems.

My web site was becoming cluttered with old, obsolete files. Also, my daily journal pages from months and years back generate a continuing series of annoying email messages from people who've found them via a search engine. I got one message the other day from someone who was annoyed because a link I'd posted for a free download of WinGate 3.0 no longer worked.

So I decided to clean things out. I deleted all of my journal pages from 2000 and prior, along with some more recent deadwood. Although these pages will no longer be available on the web, I did zip them up as a huge archive file (nearly 15 MB) and post that file to the subscribers-only area. Henceforward, only subscribers will have access to those pages, for whatever they may be worth.

I doubtless broke some links while doing this wholesale cleanup, but that's as may be. I'm mainly concerned with current stuff, and don't worry too much about older stuff. It's there, such as it is, for anyone who wants to read it, but I don't have time to maintain it.

Click here to read or post responses to this week's journal entries

Click here to read or post responses to the Linux Chronicles Forum

[Top]

Friday, 24 August 2001

[Last Week] [Monday] [Tuesday] [Wednesday] [Thursday] [Friday] [Saturday] [Sunday] [Next Week]
[Daynotes Journal Messageboard]  [HardwareGuys.com Messageboard]

09:55 - This is becoming very annoying. Roadrunner has generally provided reliable connectivity, but they seem unable to keep their mail servers up. At least once a week, it seems, their POP server and/or SMTP server goes down. The POP server being down doesn't bother me. Most mail addressed to me goes to one of my POP accounts on my server at pair Networks, so when the Roadrunner POP server is down I don't miss much. But the SMTP server is a different story.

Yesterday morning I sat there reading and answering email. When I finished, I happened to notice that my Outlook Outbox had a dozen or so messages stuck waiting to go out. I tried a couple times flushing the queue manually, but the Roadrunner SMTP server was simply dead. So I exited Outlook, told it to save the unsent messages, and tried again later. They finally did go out, or so I assumed.

Then last night I sent myself an email. I do that frequently when I need to remind myself of something. In this case, I noticed that we hadn't covered selenium in our poisons e-book, and so I sent myself an email to that effect. It never showed up in my inbox. So I sent it again, and this time it showed up immediately. That first message still hasn't showed up in my inbox, which makes me wonder how many other messages I sent are somewhere out there in digital limbo.

This morning, one of the members of our informal webring sent a message to the backchannel saying that he needed a copy of Outlook 98. I replied to that, both to the backchannel mailing list and to him privately. Both those messages were sent, but the one sent to the mailing list never showed up in my inbox.

12:03 - Thanks to everyone who's suggested bringing up a local SMTP server. I may in fact do that.

On another topic, I've finally received permission from the Zoning/Planning Board (AKA Barbara) to do something I've been wanting to do for a long time. Put a PC in the den next to the sofa. Barbara has strong opinions about where PCs belong. For example, when we were re-doing our hall bathroom, I struggled in vain to convince Barbara to let me put in an Ethernet jack next to the toilet. But this time I've presented a good case, so Barbara agrees (with reservations) that it's okay for me to put a system right next to where I spend most of my time when I'm not in my office or in bed.

I want a system there because that's where I do most of my reading. It's not convenient or comfortable to read in my office, and when I'm reading I frequently need to look something up or write a reminder note to myself. For example, I'll be reading a forensic toxicology book from 1903 and come across something I really want to add to our reference e-book on poisons. I suppose I could write myself a physical note, stick a bookmark in the book, and transcribe the stuff later, but I'm not that organized. I want to strike while the iron is hot. Similarly, I'd like to have a copy of FrontPage (we're doing the e-book in HTML) that's dedicated to the e-book. FrontPage allows me to change back and forth easily enough, but it's just easier to have different copies of FrontPage for different purposes. It would also be nice to be able to access network resources like the Internet and reference material from my seat on the sofa.

I'd been using my Compaq Armada E500 notebook for that purpose, but the problem with that is that I can't leave it running all the time. And when I put it into sleep/suspend mode, it takes quite a while to come back to life. I'm looking for something that's "instant-on" and a desktop PC best meets that criterion. Also, it's a pain in the butt to move the notebook around with an Ethernet cable and separate corded mouse attached to it. And the Compaq keyboard, while excellent for a notebook system, is not equivalent to a full-size desktop keyboard for serious typing.

So I've decided to build a desktop system and stick it next to the sofa. I'm going to base it on an old Dell Pentium/200 frame. The hard disk, CD-ROM drive, and so on are all fine as is. I'll install an Intel D815EEA motherboard with a Celeron/800 processor and a quarter-gig or so of RAM. The only thing I think I'll need to buy is a keyboard. I have tons of keyboards lying around here, but none that has a built-in mouse or trackball. Any suggestions as to the best keyboard/trackball are welcome. As far as the monitor, I'll just use an old 15" Mag-Innovision monitors I have sitting here unused. It'll run 800X600 well enough. If this works out, I'll probably upgrade to a 17" Hitachi or NEC/Mitsubishi model.

Then I need to run an Ethernet cable from the den to my office. I just checked my networking supplies. I have plenty of snap-in modular Cat-5 connectors, and free positions in both cover plates in question. I have plenty of Cat-5 Ethernet drop cables. What I lack is the 50 or 75 feet of Cat-5 or 5e cable I'll need to make the run. I have a mile or so of Cat-3, but I'd prefer not to limit myself to 10 Mb/s. So I emailed a couple of local friends to see if they have a partial box of Cat-5/5e they'd be willing to part with.

In fact, I'm debating running two cables, which would leave one spare for Barbara. No, on second thought, I'll just run one. If Barbara later needs a PC in the den, it'd be easy enough to install a small hub. And at that point, we'd be one of the few homes in the world with a separate Ethernet segment in the den.

15:06 - In case you think it only happens to you ...

As Barbara and I got the motherboard installed, the garage called to say her truck was ready. So I dropped Barbara off there to pick up her truck, after which she was heading for the library to do some volunteer work. When I returned home, I fished around until I found a Celeron/800 processor. It didn't have a heatsink/fan, so I did some more fishing around and came up with a Taisol CEK733092 (which is one of the ones I recommend). Incidentally, I really like the way Taisol HSFs snap onto the processor socket. The bracket has two holes on each side, one that fits the nubs on the socket and the other that's designed to allow you to use a screwdriver to press down until the bracket snaps into place. That was fortunate, because I'd already installed the motherboard, and there isn't a lot of room to work in the Dell case. I installed the processor and HSF, checked all the cables, connected the system to power, and fired it up.

Nothing happened. I mean nothing. No fans, no indication at all that the system was getting power. I knew this was too easy. Rather than individual two- and three-wire cables for the front panel switches and lights, the Dell case has one header-pin connector that fits the Intel D815EEA motherboard perfectly. The ribbon cable disappears inside the front panel, so I was hoping that this would be a simple matter of plugging in the cable. There's even a blocked-hole key on the connector that corresponds with a missing pin on the Intel motherboard. This should be a standard connector, but knowing Dell's tendency to "improve" things, it may not be. In fact, it probably isn't, or I would have gotten at least some indication that the motherboard was receiving power.

Oh, well. I'll figure it out this evening or tomorrow, when Barbara is here and I have more patience.

Click here to read or post responses to this week's journal entries

Click here to read or post responses to the Linux Chronicles Forum

[Top]

Saturday, 25 August 2001

[Last Week] [Monday] [Tuesday] [Wednesday] [Thursday] [Friday] [Saturday] [Sunday] [Next Week]
[Daynotes Journal Messageboard]  [HardwareGuys.com Messageboard]

08:55 - I read an article in the paper this morning about another government outrage. The INS has deported a Winston-Salem woman to Mexico. Now, it's true that this woman is not a US citizen and had entered the US illegally. But that's not the issue, or at least it shouldn't be. This woman is married to a US citizen, which in itself should guarantee her right to live in the US, and arguably should be sufficient for her to claim US citizenship. Some might argue that an illegal immigrant may marry a US citizen simply to gain US citizenship, but there is no question about this being a marriage of convenience. She has two children, both born in the US and therefore automatically US citizens, even had her husband not been a US citizen.

But the INS caught this woman at the Charlotte airport and deported her and her children to Mexico, leaving her husband alone here in Winston-Salem. Something is fatally flawed about US immigration policy when it results in young children who are US citizens either being separated from their mother or being forced to live outside the US.

I visited the Dell web site yesterday to determine the pinouts on the connector for the front panel lights and switches. I found the section for the XPS-M200s easily enough. I downloaded every document I could find, including the service manual, and nowhere could I find that connector documented. At this point, I'm not sure if I have a problem with the power supply, the motherboard, or what.

But I just noticed that I have a Duron/800 system sitting under my credenza unused. It has 128 MB of RAM, a 20 GB hard drive, and a CD-ROM drive, so it's more than sufficient for my purposes. So I think I'll fire it up, install what I need and use it in the den. I'll worry about the Dell system later.

 

Click here to read or post responses to this week's journal entries

Click here to read or post responses to the Linux Chronicles Forum

[Top]

Sunday, 26 August 2001

[Last Week] [Monday] [Tuesday] [Wednesday] [Thursday] [Friday] [Saturday] [Sunday] [Next Week]
[Daynotes Journal Messageboard]  [HardwareGuys.com Messageboard]

09:47 - After spending the day working very hard, we decided not to head up to Bullington last night to observe. The forecast was for clear, cool weather, but the humidity was to be in the mid-80% to low 90% range. That didn't bode well for observing, so we decided just to stay home and take it easy.

We got the cable run from the den to my office. I hadn't heard from either of my friends about the Cat-5 cable, so I decided to just to run Cat-3. There's not all that much real difference between 10BaseT and 100BaseT for my purposes, and the Duron machine already had a 10BaseT card installed. Running the cable took only an hour or so, and I spent much of that trying to navigate around all the stacked boxes in the attic while bent nearly double, pulling cable and stapling it to rafters. It felt like it was about 150F (65.5C) up there. Come to think of it, it probably was. Note to self: in future, avoid stringing cable in the attic in August.

It was a shame to crimp those nice Cat-5e connectors onto Cat-3 cable, but they work well enough. I tested the run by connecting my notebook, and was unsurprised to see that it linked to the hub at 100 Mb/s. A lot of people think that Cat-3 can run only 10 Mb/s and Cat-5 or better is required for 100 Mb/s, but that's not really true. Short runs of high-quality cable can easily support 100BaseT. And this was a short run, probably less than 75 feet, and it used AT&T Cat-3 cable. The limiting factor on using 100BaseT is the electricals of the cable, and short runs have better electricals than long runs.

A lot of people also think that there's a 100-metre limit on 10BaseT runs, but that's a completely bogus number. The IEEE spec says nothing about run lengths. The EIA/TIA-568 specification mentions run lengths, but its limits are 99 metres--90 metres of horizontal cable, 6 metres of cross-connect, and 3 metres of drop from the jack. Apparently, someone at some point rounded off the 99 metre 568 limit to 100 metres, and that length has become an urban legend. 

In fact, other than the quality of the physical connections, only two things matter in a cabling system: the electricals (capacitance, near-end crosstalk, and so on) of the cable and the round-trip delay (RTD). Ethernet uses CSMA/CD (carrier-sense, multiple access, collision-detect) and that CD part is important. The total length of the cabling system (or, more accurately, the RTD) has to be small enough that it is 100% certain that all collisions will be detected. If the cable is too long end-to-end, it's possible for a collision to occur at one end that won't be detected at the other end in time. But with 10BaseT, the timing is relaxed enough that lengths much greater than 100 metres are possible. At one point I ran a Cat-5 cable nearly 200 metres on a 10BaseT network, and it worked fine. Of course, 100BaseT operates at 10 times the speed, and has accordingly tighter limits on RTD.

While I was crimping on connectors, Barbara was cleaning behind the sofa and taking my end table down to bare metal. The first picture is my end table, cleaner than it will ever be again (at least until the next time...) The second picture shows where most of the stuff ended up while Barbara cleaned and lemon-oiled the end table. And the third is me with my new system installed and running.

rbt-end-table.jpg (38108 bytes) rbt-end-table-2.jpg (41751 bytes) rbt-end-table-3.jpg (38671 bytes)

At any rate, we got the cable run and everything works fine, finally. I'm writing this from my new workstation in the den. More about that tomorrow. For now, I need to get to work on the regular Sunday chores.

Click here to read or post responses to this week's journal entries

Click here to read or post responses to the Linux Chronicles Forum

[Top]

 

Copyright 1998, 1999, 2000, 2001, 2002, 2003, 2004 by Robert Bruce Thompson. All Rights Reserved.