I’ve been slower to adopt Fedora 11 and, now, Fedora 12. It took me almost 3 months, well into the Fedora 13 release cycle, to get comfortable enough with Fedora 12. Most of this was frustration with virtual machine options,  either not ready for the latest kernels (2.6.31 in Fedora 12) as with VMware Server 1.0.10, or not supporting legacy Windows (my experience with KVM and VirtualBox). Virtual satisfaction with VMware Server and kernel 2.6.31 delves into some of the details and prognosis.

The frustration led to questioning whether or how to continue with Fedora. Maybe CentOS would meet my production needs? Or wait to understand/adopt Fedora releases until the VMware community caught up with them? I was thinking a much longer wait, but the three months seems just about right this time.

After finally finding the seeming solution for VMware Server a few weeks ago, I successfully put Fedora 12 on my primary internally facing machine last week, including VMware Server 1.0.10, and transitioned the main external facing web/mail/name server to Fedora 12 two nights ago.

Outside of the virtualization issues, there’s not much to say, except that Fedora 12 seems more ready for release than Fedora 11 did. Awkward aspects of 11, e.g., needing an ext3 boot partition separate from the ext4 file systems, are gone. So far, 12 seems more polished, pleasantly surprising, and ready for prime time.

I’ve still not gotten interested in Debian proper, barely trying new releases of Ubuntu. But the comparably aggressive release schedule of Ubuntu helps. I don’t know if the virtualization issues would have been resolved nearly so quickly by the VMware community if Ubuntu 9.10 were not also using the 2.6.31 kernel. My latest understandings and successes began with reading How to install VMware Server on Ubuntu and following links and comments there.

Since the old ideas didn’t work, trying KVM seemed a likely alternative. However, my main interest in production virtual machines is running Windows 2000 Professional to run old Windows stuff that I need from time to time, stuff that isn’t likely to work with more recent versions of Windows. (I’ll probably still want Win2K for these purposes even when the support life cycle ends July 13. I still find use for NT4 Server for analogous reasons.) I don’t know whether the KVM project even intends to support legacy Windows. In any case, the Win2K install process seemed to get started OK, but then would not complete. It has been a couple of months since I tried this, so I don’t remember all the details.

While looking for solutions, I kept seeing praise for VirtualBox. The VirtualBox documentation showed promise, particularly the support for existing VMware partitions. I installed an rpm, tried to fire up an existing Win2K bootable partition, but the boot process didn’t get very far. I didn’t succeed in installing Win2K with VirtualBox, either.

Sometimes the right solution is to wait. Roughly once a week, or so,  I’d search to see if there were new possibilities emerging.

About a month ago, I discovered How to install VMware Server on Ubuntu. Though not of direct help, it did link to Kang’s vmware-specific-specific 5.5.x and kernel 2.6.31. I’d seen and used some of Kang’s VMware Server patches before, but the comments on that page referred to init_mm problems that I didn’t know how to solve for Fedora 12.

Aack’s comment #32, posted on January 13, provided the simplest solution I’ve seen to the init_mm problem, simply replacing one line in pgtable.h to avoid the problem altogether. That change doesn’t require kernel builds, it gets picked up when vmware-config.pl builds modules specific to the running kernel. I didn’t notice the comment for some days after it was posted, and also wanted to do some disk reconfiguration before testing, but once I got to testing with the one line change, all seemed well.

So thanks to Kang and Aack, I’m running VMware Server 1.0.10 on Fedora 12 with no apparent difficulties.  Kang’s comment #31, just before Aack’s praised VirtualBox, but Kang still seems to be preparing for kernel 2.6.32 (vmware-specific-specific 5.5.x and kernel 2.6.32), so maybe things will be easy enough by the time Fedora 13 comes out.

My skepticism dates back to first reading some of the patents in 2004 and the press releases Avistar produced regarding Microsoft’s requests for re-examination of some of the patents (see public display of dis-affection: Avistar patents & Microsoft and AVSR vs. MSFT: numbers of patents or patent numbers?).

In hindsight, it seems that Avistar chose to change course about the time that it brought Bob Kirk in as CEO last July. After that time, Avistar said little publicly about the patents, seemingly not saying anything about U.S. PTO actions or licensing progress. January 19, two days before the Intellectual Ventures transaction, Avistar sold a license to Skype for $3M.  If this had happened a couple of years ago, there would likely have been significant publicity. The only Avistar publicity about the Skype transaction appears to be the SEC 8-K filing.

In the January 26 press release, Bob Kirk is quoted as saying “This also removes what has been a significant distraction for the team and our operation in general, while providing us the capital to more aggressively invest in our business.”

It is difficult to say more without speculation. Avistar had announced licenses with many potential licensees, but, as far as I can tell, had not succeeded in negotiating licenses with Cisco, Hewlett-Packard or Microsoft — it is hard to imagine easy negotiations with those companies. I find it easier to imagine that those companies are some of the investors in Intellectual Ventures and receptive to this latest news. [Update February 26: Verizon Patent Case Marks a First for Intellectual Ventures lists both Cisco and Microsoft as Intellectual Ventures "members".]

Along with the promise have been ambiguity, questions and controversy. This is a brief recap while still waiting for some of the partners to make their moves.

For example, when Cisco announced plans to purchase Tandberg, the largest of the companies committed to ITU-T and SIP interoperability, would Cisco become part of the interoperable crowd, or would Tandberg become less interoperable? It seemed inevitable that interoperability would prevail, but until it happened, who could be sure? Then the real controversy emerged: enough Tandberg shareholders want a higher bid from Cisco that the deal may not happen. A November 9 deadline has been extended to November 18.

The long anticipated sale of Skype by eBay was announced in September in the midst of controversy over intellectual property, and prior announcements of Asterisk interoperability and SIP interoperability. Naysayers widely predicted the demise of Skype. Others, notably LifeSize, joined the dance. With Friday’s announcements resolving the ominous litigation, Skype’s forward momentum seems impressive. Skype hiring of SIP pioneer Jonathan Rosenberg bodes very well for future Skype interoperability.

Assuming Skype overcomes current lack of multi-point video calling, Skype should be able to win a few dance contests.

Other recipes I’ve found don’t work, and at first I couldn’t get this one to work because it requires building a kernel from source.

The change to the kernel source is only one line, a directive at that, so the source changes aren’t intimidating, but the kernel build and install process are intimidating, at least for someone like me who last built a kernel in 1989 (BSD 4.2 if I recall correctly). Things were certainly simpler then, but things we take for granted today, like VMware, weren’t available, either.

I followed the instructions at Building a custom kernel – FedoraProject first with 2.6.29.6-213. That never worked. I’m not sure if there was a problem in the source RPM or, more likely, that I skipped a critical step. I was using very slow, old iron, a 733MHz Pentium III with only 512M of physical memory.

Busy with other things and frustrated with the process, I let that machine sit until Fedora 11 updated to 2.6.29.6-217.2.3. Then I tried again, followed the instructions more carefully, pruned away unnecessary processes that were gobbling memory, and got a successful build. Successful builds took about 10 hours. I had neglected to build kernel-firmware the first time so repeated the process with better build parameters.

After installing the kernel and firmware RPMs and rebooting, things were ready for the VMware install and configuration. The main remaining glitch was that installing the kernel had not setup paths that vmware-config.pl is used to looking for, mainly /lib/modules stuff. With a kludgey symlink, that seems adequately resolved for making VMware Server work OK. (I have a Fedora 1 virtual machine running on that underpowered hardware, seemingly running OK.)

However, my kludges don’t seem weren’t sufficient to make Asterisk (really DAHDI) build and install properly. So I need to better understand how to get them going. I have a guess, but need to follow up. Update August 16: All that seems needed is this symlink:

ln -s ~/rpmbuild/BUILD/kernel-2.6.29/linux-2.6.29.i686/ /usr/src/kernels/`uname -r`

The equivalent of that probably gets instantiated by the kernel-headers RPM, but I didn’t think to build that one.

Most people don’t have a use for hardware this old, but I find myself accumulating machines to use as spares, in case the original Optiplex gives out, and for testing and random experiments. I wouldn’t run Windows on this hardware, and probably wouldn’t tolerate the performance of GNOME or KDE on an ongoing basis, but  for “server” stuff it is quite serviceable.

The biggest challenge is memory consumption. This GX1 has 3 slots, and there are 512M DIMMs that would fit physically, but the circuitry won’t address more than 768M. I have a couple of Optiplex GX100s with faster processors, but only 2 memory slots, so they’re maxed out at 512M. They’re useful for testing and one stands ready as a hot spare for this machine, but the difference between 512M and 768M is quite noticeable.

The 512M constraint was most noticeable building a kernel from source. To avoid severe swapping, or worse, I had to stop all the GUI stuff, MySQL, httpd, etc. Even then, watching the memory consumption made me wonder if the build would complete. It did, but it took about 10 hours on a 733MHz Pentium III! The corresponding build on a 2.8GHz P4 with 2G of memory took about 2 hours, a factor of 5 improvement vs. the 3.8x improvement one might expect based on clock frequency. Even that comparison is probably understated, since the 2.8GHz/2G machine was still runing the GUI stuff, MySQL, httpd et al during the build.

Oops. I really meant Apollo 11 delivered…

But it’s not July 20 anymore, so about Fedora 11:

• Overall, no big problems
• Fedora Project slipped their final release schedule a couple of weeks, so I didn’t get started trying Fedora 11 until mid-June.
• VMware Server 1.0.x still doesn’t work with the 2.6.29 kernel(s) in Fedora 11. It appears that a one line kernel change is needed (assuming VMware doesn’t fix directly). However, I’ve never built a linux kernel before, and my first attempts have failed.
• The nastiest surprise, for me, was confusion about BIND. I’m used to Fedora putting BIND in a chroot’d jail. Fedora 11 seems to eschew actually doing this, but provides the /var/named/chroot directory hierarchy as if the jail still exists. I don’t find anything in the release notes about any of the BIND changes, and the additional DNSSEC  support in BIND 9.6 threw me off temporatily, since I don’t know much about DNSSEC. It took me a couple of hours to sort everything out, and my current solution is a bit clumsy, but seems to work.

There are other awkward aspects, such as the need for a /boot ext3 partition when trying to use ext4 for the rest of the filesystems, but these are adequately documented in the release notes, so not big problems for me.

I put Fedora 11 on my primary mail/web/DNS server yesterday, and all seems OK so far. (This post is stored on that server.) But the machine that depends on VMware Server is still running Fedora 10.

Red Arco Iris (en Inglés, Rainbow Network) is coming of age and has much to celebrate. From Keith and Karen Jaspers’ vision 15 years ago to today, much has been accomplished.

Outside of Ciudad Sandino, in Comarca: Trinidad Central, last week we celebrated the 25 millionth meal served at our feeding centers. The rate of malnutrition in the 124 communities we serve has gone down from about 50% upon our arrival to less than 25% today. [see also Aleyda Hernandez Tells Her Story]
Also outside of Ciudad Sandino, Housing Project Cuajachillo 2 celebrated the completion of the payoff of the (ten-year) mortgages for the houses Red Arco Iris built there. One of those celebrating was the baby girl in the center of the photo above. She is now the pre-adolescent with Keith and her baby photo, at right.
Health care miracles abound, e.g., Victoria walking again after a dozen years of immobility.
Students have progressed from our elementary schools, to sponsorship in high school, on to further education and/or to establishing businesses, such as that of the poultry entrepreneur in San Isidro, outside of Nagarote.

Of course, there is much more ahead in service, in internal adjustments as leaders step back and step up, and in partnership with those of like minds. I’m particularly intrigued by possibilities with el Porvenir and with el Samaritano.

I spent Friday morning with the managers of the seven Red Arco Iris project regions (“networks”) and other Nicaraguan staff, working towards la red de redes (the network of networks). I believe, and I think the staff believes, that improved use of email, use of Internet shared storage (SkyDrive) and Skype calling will facilitate better communication and more efficient use of time and other resources.

One of the surprises of that session was eagerness to try Linux. I’m encouraging exploration using Ubuntu “live” CDs.