Much of my time is focused on helping the poorest of the poor in rural Nicaragua, where families subsist on US $1/day income. Red Arco Iris, as it is known in Nicaragua, a.k.a. Rainbow Network in the U.S., provides feeding centers, micro-loans, housing assistance, schools and educational assistance, and physicians and dentists at community clinics.

Of course, computer technology, Internet communication, cell phones, fax machines and other standard office accoutrements enable and facilitate Red Arco Iris staff in the seven regional offices. The staff enablement may be felt indirectly in the communities, but we can be much more ambitious.

Our ”feet on the ground” efforts have little to do with modern technology: food at the feeding centers is prepared on open fires, micro-loan committees depend on hand-writted communication, the houses we help build are concrete/cinder-block/tin-roof construction, classes may be held under a shade tree, doctors practice in makeshift clinics, and so forth.

However, modern technology “inevitably” finds its way into direct import in these rural communities, whether it be a television hooked up to hazardous power or a cell phone passed amongst the people. Portable computers are beginning to be present. Internet connections are plausible. 20-year-olds who ten years ago couldn’t attend school are now graduating from high school and going on to college, where the technology needs and wants are not hard to imagine.

Assuming technology incursions are inevitable, are on the cusp of meaningfulness in rural Nicaragua, the questions become “How can technology be positive in these communities?” (money going to the cell phone providers is positive!?) and “How can we facilitate positive technology infusion?”.

Lots of small steps are being made. In days to come, there will be more to say about those steps.

Before placing the order, I wanted to be sure that dual booting would be feasible. I found instructions: How to dual boot Vista and XP (with Vista installed first) that made sense. When I got the machine, I followed those instructions, and they seemed to work. However, I kept fiddling around and breaking either Vista or XP or both, and ended up installing Vista again, as well as XP.

Once beyond those basics, everything went smoothly. I’ve been pleasantly surprised to find that everything I want to do with the machine, so far, works at least as well with Vista as with XP. In a few cases I’ve had to download Vista-specific device drivers, but that much bother was the minimum to be expected.

Running Vista locally, instead of remotely with VNC, has alleviated my minor grumblings about my earlier Vista experience. I have much of Aero turned off. User Account Control is still a little clumsier than it should be, but so be it.

I’ve been using the machine seriously for about 10 days. Overall, I’m delighted with it, but that is probably more due to the hardware than Vista. But, counter to my expectations, I’m not using XP at all on the new machine, except to test things.

Some observations based on numerous experiments with combinations of NT4, Windows 2000, Server 2003, Server 2008 and available service packs:

1. Microsoft could make it easier to upgrade by saying explicitly in an obvious place, something like “Unlike Windows Server 2003, which allows Windows NT Server machines to co-exist as domain controllers, Active Directory in Windows Server 2008 requires that domains be at least at the Windows 2000 native level.” Perhaps this is said some where more explicitly and obviously than in the error messages from running “adprep.exe /domainPrep”, but I haven’t found that place.
2. Microsoft could make it easier by allowing a direct upgrade path from NT4 to Windows Server 2008. Direct upgrade from NT4 to Server 2003 is possible — why not preserve this ability? At least two possible answers: NT4 systems are likely to have at most 4GB primary partitions, and 2008 requires 6GB in the primary partition. (Most NT4 system install discs were at the NT4 SP1 level, and those discs wouldn’t create a primary partition larger than 4GB.) Also, the device driver model of 2008 is enough different that older devices may not have compatible drivers. Though tedious, it seems the easiest (in-place) upgrade path from NT4 to Server 2008 is to go through both Windows 2000 Server and Server 2003 as intermediate steps. After being stymied by device driver issues, I eschewed in place upgrade and now have test machines with all four editions (NT4 SP6, 2000 SP4, 2003 R2 & 2008).
3. With all of the un-installs/re-installs I anticipated, I didn’t want to complicate things by involving Windows DNS. Though “everyone” discourages using BIND with Active Directory, I found it painless to use BIND as long as I put the right SRV records in the zone file. (I used the records described in Bob Carver’s Configuring Berkeley Internet Name Domain (BIND) to Support Active Directory.)
4. With the exception of an old notebook that was already running NT4 Server SP6, and an XP client for testing, all of these experiments were in virtual machines (VMware Server 1.0.6, mostly hosted by XP Pro plus some instances of Fedora 9 hosting). The experiments were relatively painless that way, but would have been foreboding if done directly on real machines.

So now I too still ponder saying, regarding upgrading from NT4 Server to Server 2008, “We could do it … but it would be wrong”.

The robustness of NT4 SP6 was a pleasant surprise. Though I prepared to re-install NT4 as experiments failed and AD changes were made, NT4 just kept on ticking as if it hadn’t even taken a licking. And NT4 domain controllers inexplicably seemed to co-exist with AD even after AD went to native mode.

There’s still nothing so broken the way we’re using NT4 Server that we need to upgrade. We don’t come close to the thousands of accounts limits of NT domains. There are Vista drivers for the “big printer” cited in the May 12 post, drivers that seem to use LPR and not involve the NT4 Server. Since that was the biggest existing problem in our NT4 Server environment, there’s still little motivation to change. And, if printing isn’t an issue, maybe Samba would be sufficient if all we really need is a file server??

Update July 5: see Why I Still Use Windows 95

Some observations based on numerous experiments with combinations of NT4, Windows 2000, Server 2003, Server 2008 and available service packs:

1. Microsoft could make it easier to upgrade by saying explicitly in an obvious place, something like “Unlike Windows Server 2003, which allows Windows NT Server machines to co-exist as domain controllers, Active Directory in Windows Server 2008 requires that domains be at least at the Windows 2000 native level.” Perhaps this is said some where more explicitly and obviously than in the error messages from running “adprep.exe /domainPrep”, but I haven’t found that place.
2. Microsoft could make it easier by allowing a direct upgrade path from NT4 to Windows Server 2008. Direct upgrade from NT4 to Server 2003 is possible — why not preserve this ability? At least two possible answers: NT4 systems are likely to have at most 4GB primary partitions, and 2008 requires 6GB in the primary partition. (Most NT4 system install discs were at the NT4 SP1 level, and those discs wouldn’t create a primary partition larger than 4GB.) Also, the device driver model of 2008 is enough different that older devices may not have compatible drivers. Though tedious, it seems the easiest (in-place) upgrade path from NT4 to Server 2008 is to go through both Windows 2000 Server and Server 2003 as intermediate steps. After being stymied by device driver issues, I eschewed in place upgrade and now have test machines with all four editions (NT4 SP6, 2000 SP4, 2003 R2 & 2008).
3. With all of the un-installs/re-installs I anticipated, I didn’t want to complicate things by involving Windows DNS. Though “everyone” discourages using BIND with Active Directory, I found it painless to use BIND as long as I put the right SRV records in the zone file. (I used the records described in Bob Carver’s Configuring Berkeley Internet Name Domain (BIND) to Support Active Directory.)
4. With the exception of an old notebook that was already running NT4 Server SP6, and an XP client for testing, all of these experiments were in virtual machines (VMware Server 1.0.6, mostly hosted by XP Pro plus some instances of Fedora 9 hosting). The experiments were relatively painless that way, but would have been foreboding if done directly on real machines.

So now I too still ponder saying, regarding upgrading from NT4 Server to Server 2008, “We could do it … but it would be wrong”.

The robustness of NT4 SP6 was a pleasant surprise. Though I prepared to re-install NT4 as experiments failed and AD changes were made, NT4 just kept on ticking as if it hadn’t even taken a licking. And NT4 domain controllers inexplicably seemed to co-exist with AD even after AD went to native mode.

There’s still nothing so broken the way we’re using NT4 Server that we need to upgrade. We don’t come close to the thousands of accounts limits of NT domains. There are Vista drivers for the “big printer” cited in the May 12 post, drivers that seem to use LPR and not involve the NT4 Server. Since that was the biggest existing problem in our NT4 Server environment, there’s still little motivation to change. And, if printing isn’t an issue, maybe Samba would be sufficient if all we really need is a file server??

Update July 5: see Why I Still Use Windows 95

I’m leery of the clerical accuracy of the “Patent Application Information Retrieval” (”Public PAIR”) data at http://portal.uspto.gov/external/portal/pair, because of typos and worse that I have found in the past/find there now, but I have looked there at a few of the Avistar patents. There seem to be three patents of special interest when Avistar asserts patents against other companies. For example, from Avistar Announces Patent Infringement Lawsuit Against TANDBERG:

“Avistar Communications Corporation (Nasdaq:AVSR), a provider of business video collaboration solutions and technology, today announced that its wholly owned subsidiary, Collaboration Properties, Inc. (CPI), has filed a patent infringement lawsuit against TANDBERG ASA and TANDBERG, Inc. In the suit, CPI alleges that several TANDBERG videoconferencing products infringe three of CPI’s patents. The three patents involved are U.S. Patent No. 5,867,654; No. 5,896,500; and No. 6,212,547.”

According to PAIR:

• Patent number 5,867,654, “Two monitor videoconferencing hardware”, has a status of “Reexam Assigned to Examiner for Determination” as of April 11. (Update June 14: ‘654 request for re-exam is now listed as denied, as of June 10.)
• Patent number 5,896,500, “System for call request which results in first and second call handle defining call state consisting of active or hold for its respective AV device” has a status of “Request for Reexamination Denied” as of May 30.
• Patent number 6,212,547, “UTP based video and data conferencing”, has a status of “Determination - Reexamination Ordered” as of May 20.

So for these three main patents, the score seems to be Avistar 1 (#500) , Microsoft 1 (#547), with the tie breaker (#654) pending.

But I suspect that the number of patents requested for re-exam, and the numbers of patents where Avistar and Microsoft win these preliminary determinations, will matter very little relative to what happens with that one patent where reexam is ordered.

6,212,547 is likely more important than any of the other numbers above.

Update: June 14: The problem was not limited to Fedora. The problem also occurred with Windows Server 2003 in a virtual machine. The problem was the free version of ZoneAlarm. Uninstalling ZoneAlarm and using only Microsoft’s firewall, routing works OK in the previously problematic environment. ?? It is hard to identify a meaningful environmental difference between the unsuccessful test deployment vs. a separate successful test deployment vs. the production environment. All three physical machines have Broadcom 440x 10/100 Ethernet controller, so differences in the network driver paths, my first hypothesis, seems unfounded. The unsuccessful environment has XP Pro as the real OS, while the successful environments have XP Home. ?? 

Fedora 9 GUI stuff still seems less stable:

1. Every time I’d start VNC, I’d get a kernel oops warning, but was unable to get the applet to report the oops, so I turned off the warnings.
2. The updates pending alert has never yet allowed me to apply an update
3. Many times when I try to start a GUI applet (application??) I get a big warning box “There was an error starting the GNOME Settings Daemon.” There’s no indication of any way to try to fix, just “GNOME will try to restart the Settings Daemon next time you log in.”
4. System->Administration->Update System, like the automated alert, does not allow me to apply updates.
5. System->Administration->Add/Remove Software never lets me add software. So I can’t say anything good about PackageKit in Fedora 9.

I routinely used pirut, pup (& puplet) in the past, but since those seem to have been removed, I use yum from the command line. I’ve grown to mostly like that better than the older GUI stuff, but still found those useful enough to wish they were still around.

Since I mostly use Fedora for name/mail/web serving and similar things, none of these problems are a big deal for me. But they are annoyances.

I’ll probably start putting Fedora 9 into production on my two main servers after a new disk arrives tomorrow.

It is traditional and usually judicious to be leery at first when Microsoft releases a new version of Windows (or any software supplier releases a new major version of any piece of software). Even Windows 3 (3.1?) wasn’t really ready until Windows 3.1 (3.11?) Windows 98 might have been OK the first time out, but making it so was a big priority at Microsoft. NT4, Windows 2000 and Windows XP all needed at least one service pack before I was ready to trust them. (NT 3.1 and 3.5 were somewhat different in that they weren’t really intended for regular end users. Windows ME was a disaster, from many perspectives, and doesn’t fit in this context.)

So I barely touched Vista last year. Now Vista SP1 is the norm. From my recent experience, Vista SP1 is certainly not terrible, and seems to have noticeable security improvements over XP.

A couple of weeks I needed to purchase a new computer for my church library. After shopping around online and visting Fry’s a couple of times, the best hardware bang for the buck seemed to be a Compaq SR5410F at Circuit City. That machine came only with Vista SP1. Some of the other possibilities were available with XP, but it seemed time to give Vista a chance.

So far, everyone who has used this machine, seems quite happy with it (myself included).

All of the escalation of privilege dialogs to do administrative things are understandably annoying, but most of them seem sensible. The question is not “Why is Microsoft forcing this on users?” but rather “Why did Microsoft take so long to follow the lead of other operating systems in this regard?”.

I could do without the “Aero” 3D dialog eye-candy, but I’ve felt that way about most of analogous stuff in many prior editions of Windows, Mac OS, and the prominent X-Windows managers.

My biggest frustrations are with ancient applications that don’t work with Vista, but the frustration is best placed with the vendors who sell those applications. The most annoying instance is from a vendor whose site only mentions Windows 9X and NT, while still selling support subscriptions.

The only other problem that comes to mind is that TightVNC doesn’t work with Vista yet. But UltraVNC mostly does work with Vista. The only problem I’ve noticed with UltraVNC is that the option to turn off Aero is unreliable. When Aero crops up remotely, frequent screen refreshes are necessary. Not a big deal.

I probably wouldn’t want Vista on older hardware, but on new hardware it seems to make sense.

May 28: Slashdot seemingly soothes some more of the Vista uproar in Windows 7 Won’t Have Compact “MinWin” Kernel.

Disclaimer: I still own shares of MSFT.

Yesterday I mixed the audio, used it in place of the camera’s audio track, and broke it up into segments for YouTube: Charlie Abbitt - Live at The Wellington May 22, 2008.

I’d hardly touched Cubase LE since August, and didn’t really remember much of what I had learned back then. What I did remember was sending analog output from the iO|14 to the audio input of another computer for mixdown. Though that memory was correct, that was not the best approach.

Fortuanately, I have been using Audacity frequently for simpler recording and sound processing. I routinely export processed audio from Audacity without involving a second system. My main learning yesterday was that Cubase does have export facilities (D’oh!), whch are quite usable and useful, such that a separate computer for mixdown was unnecessary (and potentially would have compromised audio quality).

Before the realization that a second computer was wrong headed, I thought about using the recently acquired iMac G4 pedestal. I thought I spied an analog input on the back next to the (”headphone”) analog output, but I was wrong — that connector is an Apple proprietary “Apple speaker minijack  for connection to Apple Pro Speakers”. OS X System Profiler and System Preferences tell me the only built-in audio input is the mic. Determining that was not easy, with only minimal info readily available from both apple.com & Google, but one of my Mac expert friends confirmed my determination. Even he had to stare at the back of one of his pedestals to be sure.

To try to end this rambling, but finish the story, this morning I wondered more about the 1/8″ combined analog/digital audio jacks on more recent Macs. The iO|14 has S/PDIF coaxial RCA digital input/output. The Mac jacks are S/PDIF optical, but not the typical TOSLINK, since those are different size/shape from the ubiquitous 1/8″ analog jacks. I finally found the explanation that the Mac jacks are “mini-TOSLINK”, that optical adapters are inexpensive, and that bi-directional coaxial/optical converters are not expensive. 

Yesterday I mixed the audio, used it in place of the camera’s audio track, and broke it up into segments for YouTube: Charlie Abbitt - Live at The Wellington May 22, 2008.

I’d hardly touched Cubase LE since August, and didn’t really remember much of what I had learned back then. What I did remember was sending analog output from the iO|14 to the audio input of another computer for mixdown. Though that memory was correct, that was not the best approach.

Fortuanately, I have been using Audacity frequently for simpler recording and sound processing. I routinely export processed audio from Audacity without involving a second system. My main learning yesterday was that Cubase does have export facilities (D’oh!), whch are quite usable and useful, such that a separate computer for mixdown was unnecessary (and potentially would have compromised audio quality).

Before the realization that a second computer was wrong headed, I thought about using the recently acquired iMac G4 pedestal. I thought I spied an analog input on the back next to the (”headphone”) analog output, but I was wrong — that connector is an Apple proprietary “Apple speaker minijack  for connection to Apple Pro Speakers”. OS X System Profiler and System Preferences tell me the only built-in audio input is the mic. Determining that was not easy, with only minimal info readily available from both apple.com & Google, but one of my Mac expert friends confirmed my determination. Even he had to stare at the back of one of his pedestals to be sure.

To try to end this rambling, but finish the story, this morning I wondered more about the 1/8″ combined analog/digital audio jacks on more recent Macs. The iO|14 has S/PDIF coaxial RCA digital input/output. The Mac jacks are S/PDIF optical, but not the typical TOSLINK, since those are different size/shape from the ubiquitous 1/8″ analog jacks. I finally found the explanation that the Mac jacks are “mini-TOSLINK”, that optical adapters are inexpensive, and that bi-directional coaxial/optical converters are not expensive. 

Fedora is my preferred *nix platform for lots of “server-ish” things: mail, web, DNS, MySQL, Asterisk, … Linux, specifically Fedora, usually handles these chores well on relatively ancient, low-performance hardware. For a production externally facing server, I use a 450 MHz Pentium II with 768MB, and it usually idles away. I also use Fedora on a 3GHz P4 with 2GB of memory for various production purposes, including running Windows 2000 on VMware Server 1.0.5. All those things are going smoothly with Fedora 8. I also manage a couple of other machines that have Fedora 7 as a virtual machine on VMware Server on Windows XP. I’m anticipating moving all of those to Fedora 9, but having second thoughts.

First step in how I try a new edition is an install on a barely used 450 MHz Pentium II machine (512MB), to see what is changed and what administrative adjustments need to be made. Usually this is simply tweaking my scripts that work on the current edition. However, the basic network install was scrambled in the Fedora 9 install. /etc/sysconfig/ifcfg-eth0 had new fields intended for compatibility with Network Manager. These extra fields seemed to be wrong. In any case, routing outside the local subnet and DNS, even to a name server on the local subnet, did not work. Eventually I figured out how to fill in the new fields so that networking was normal and so that Network Manager would work as well. But other traditional files, e.g., /etc/resolv.conf for listing name servers, seem to have been needlessly deprecated for sake of putting similar fields in the icfg-* files.

This also seems to be yet another instance of where GUI administration breaks non-GUI administration. Microsoft seems to have figured out you need to have both, as evidenced by the “Server Core” non-GUI subset of Windows Server 2008. Why can’t Fedora (really, Linux), preserve non-GUI administration while adding GUI stuff?

Otherwise, the problematic changes from Fedora 8 seem pretty minor and easy to adjust to.

Second install test is to try a similar install in a VMware Server (1.0.5) virtual machine. So far, I have not been able to figure out to fix ifcfg-eth0 so that routing works. That makes Fedora 9 unusable in that enviroment.

I’ve also tried the ”Live” (bootable CD) option. On older hardware, Live seems fine.  In a VMware Server virtual machine, the Network Manager routing problems are present with Live and make it almost useless. On a brand new Compaq SR5450F with an NVIDIA GeForce 7100 display controller, Fedora 9 doesn’t seem to have a useful driver and punts to text-only mode.

I imagine I’ll piddle around with Fedora 9 more and deploy on the real hardware. But unless I can find a resolution for routing in the VMware virtual machines, I’ll be stuck with older editions there.