Lot 49: Greg Beaver's blog

As of a few minutes ago, I finally released:

1. PEAR 1.4.0b1 - the first beta release of PEAR 1.4.0, feature frozen. This version passes all tests, and there is only one known issue having to do with a BC break in the XML parser in PHP 5.1
2. PEAR_RemoteInstaller 0.1.0 - this is a breakaway of the remote FTP installation procedure from the PEAR base package
3. PEAR_PackageFileManager 1.6.0a2 - fixes several issues discovered in 1.6.0a1
4. PHP_Archive 0.6.0 - full PHP 5.1+ requirement, streamlining and unit testing, much more functional

Those of you following the PEAR_ErrorStack issue may notice that it has not been split off from the rest of the package. This is because of a few rare reports of PEAR/ErrorStack.php not being installed with no error reported. This issue concerns me quite a bit because in this situation, the pear command would cease to function, which is a Bad Thing(TM).

Until this issue's cause it determined, it will not split off into its own package. I actually never could reproduce the problem, but one user managed to duplicate on unix and windows within 10 minutes doing the exact install sequence I asked him to do, so who knows.

This release of PEAR is a big deal. Feature-frozen PEAR means that after some time is allowed to fix any minor or major issues encountered, PEAR 1.4.0 stable is on the horizon.

Finally, the North American Van Lines company reps pulled up this afternoon to unload my stuff. They were a mere 16 days past their original final delivery estimated date (which was 9 days later than we had requested in the first place...) However, the good news is that the only thing which appears to be missing is a small box. I may even be able to repair the damage to my laptop, which completely crashed yesterday and won't load at all because the codepage and a number of .dlls are corrupted (ah, windows, how we love thee).

The moral of the story here is pretty obvious: don't use a moving company if you

1. want your stuff to arrive on time
2. want your stuff to arrive

Recently, the news has been all a-tingle in the USA with the president's announcement that he believes Intelligent Design should be taught alongside evolution in America's science classrooms. I actually think this would be great. However, I don't think proponents of this policy really understand what this would mean.

Remember: science is a discipline based upon the "scientific method." What this means is remarkably simple in principle:

1. observe data
2. seek patterns
3. make conjectures or predictions of future behavior based on the patterns (hypotheses)
4. experiment to test hypotheses
5. adjust hypotheses based on observed results
6. "upgrade" hypothesis to theory if enough time has passed of no conflicting data with hyphothesis
7. intrinsic uncertainty of human capacity and understanding means theories can never be conclusively proven, only disproven.

The most obvious similarity in all of these steps is that observation comes *first*, then followed by hypothesis. A hypothesis that does not explain all of the data is by definition false and must be modified or thrown out. Point #7 is something that creationists consistently misrepresent, saying that something which cannot be proven is false. This is not the case, as can be shown with a simple thought experiment:

The majority of Americans have never been to Africa. If you have not been there, you cannot prove it actually exists. This does not mean that Africa is false, or even a figment of some vast conspiracy's imagination, just that it cannot be proven to exist without more data (i.e. what you see after stepping off of a plane/ship). However, all evidence supports the existence of Africa, even if you haven't been there (TV shows on African wildlife are consistent, people you've met claiming to be from Africa have similar features/languages, and so on).

However, once a hypothesis is proven false (such as "The Earth is flat"), it is no longer valid and must be then dropped from consideration as a scientific hyphothesis.

Following this train of thought, if you examine creationism (under any name) and make predictions and test them experimentally, there are a number of inconsistencies with the hypothesis and evidence that even a high school science class could verify. This would be a wonderful exercise in applied research and an excellent way to teach science to children. Unfortunately, in the process, it would disprove the scientific merit of Creationism (read: prove it false) and shake the foundation of some people's faith at the core. I would hate to shatter faith like that, but I am also confident that most of these people would lift themselves up, dust off, and find that their faith is indeed not based on little details like whether a few words written thousands of years ago (and translated recently) are 100% fact. No, I suspect they would find that their faith is based upon principles such as humility, supporting others in times of trouble, and on doing good works. It may even be based upon such radical ideas as singling out the poor, the imprisoned and the diseased to offer worldly possessions and moral support. It might even be based on a certainty of unconditional salvation that does not need to ignore the world as it truly exists.

Or perhaps, as in the words of Theodoric of York, Medieval Barber: "Nahhhhh!"

Just a few minutes ago, I was attempting to downgrade my existing PEAR to an earlier version so that I might be able to test a few things. So, via

$ pear up -f PEAR

I figured I would get the 1.4.0a12. Unfortunately, I forgot that my preferred_state was set to "stable" and ignored the error message, trying again. Halfway through, I realized my mistake - I was upgrading to 1.3.6, which would bork everything. So, I reflexively hit Ctrl-C. Uh-oh. A quick peek at the disk revealed that I had hosed the entire installation . Thinking this was going to be a bit harder to fix, I suddenly remembered that I have a convenient go-pear.phar sitting there that will install PEAR 1.4.0b1, so I did this.

It worked great, except instead of installing into C:\Program Files\php\PEAR as I thought I asked it to do, it instead installed into C:\Program Files\php\PEAR\PEAR. Crap. After a bit of hand-deleting and a re-run, I have my PEAR installation back again.

I guess the moral of the story is it doesn't matter how experienced you are with a program, read the damn error messages and prompts!

As of today, PEAR 1.4.0 (HEAD of CVS) has feature-frozen. This means all development leading up to and after PEAR 1.4.0 will only contain bugfixes. Also, development of go-pear.phar has reached a great moment: it actually works. In addition, instead of being located at pear.chiaraquartet.net, you can grab the latest version from CVS of cvs.php.net/pear-core/go-pear.phar via anonymous checkout (use google to find instructions if you're new to cvs.php.net)

Also new, the remote installation feature (installing through FTP) has been split into its own package, probably PEAR_RemoteInstaller, although approval is pending at pear.php.net. Once things are settled, expect PEAR 1.4.0b1 any day now. That's right, beta. PEAR 1.4.0 stable is not too far in the future either!

Grab the latest go-pear.phar (600k) and start installing PEAR successfully on unix and windows!

After several weeks of wrangling with the differences in streams support between PHP 4.3, 5.0, and 5.1, Davey and I have decided to drop support for PHP versions earlier than 5.1. Normally, I am all in favor of supporting the earliest version possible, but in this case appearances were deceiving. Streams support is innovative and wonderful, but simply doesn't work consistently between OSes and across versions until PHP 5.1.

For instance, PHP 4.3 does not have support for url_stat() or opendir functionality. PHP 5.0 has a buggy stream_eof() implementation and so on and so on.

More importantly, the __HALT_COMPILER() directive is available in PHP 5.1, allowing full utilization of binary to compress files. This reduces go-pear.phar from 814k down to only 600k, a significant drop. In addition, the code is safer because all internal stat caching is performed in a private static class variable instead of a global variable. It's just good all around.

So, PHP_Archive 0.6.0 will have support for PHP 5.1.0b1 and newer - and it is fabulous!

A big thanks to Wez and Sara for coding and maintaining streams. An even bigger thanks to the PHP crew for developing 5.1. Even with all of the forced changes in references, the work required to get code up to snuff is far outweighed by the rewards. I can't wait to see 5.1 stable out there.

As of a few minutes ago, I finished the sync of pearweb's REST features with peclweb. After running the generation script on the server, pecl.php.net is fully REST-enabled. This is in keeping with the push to remove all xml-rpc from the PEAR installer brought up by recent security vulnerabilities in XML_RPC. This was one of the last steps prior to feature-freeze for the PEAR package. Once binary package install has been thoroughly reviewed, we will be ready to step up to beta stability. All that is needed is a few more unit tests and complete API docs. PEAR 1.4.0 is coming!

Thanks to the brilliant organization of North American Van Lines, my desktop (linux) is a month late in delivery to my new locale in Lincoln, Nebraska. More on those little f$!+/ers later. Basically this means my quick creation of API docs for PEAR is delayed substantially. It also means basic living is very difficult as well, so please be patient if you are working with me on projects. Fortunately, I have my laptop (windows) which has made development possible. I am currently working on the last stages of fixing issues with REST caching, enabling support for REST at peclweb, and getting go-pear.phar to work properly in PHP 5.1. More on this news later.